Cellular cartography: Towards an atlas of the neuronal microtubule cytoskeleton

Introduction

From the mistaken identification of microtubules in electron micrographs as “long, tubular elements of the endoplasmic reticulum” in the dendrites of neurons (Palay, 1956), to the high-resolution studies of microtubules presently being conducted, the scientific exploration of microtubules has long been a visual one; advances in techniques have furthered our understanding of microtubule diversity, organization, and function, and reciprocally, our desire to further this understanding has driven technological advancements.

Shortly after the naming of microtubules (Ledbetter and Porter, 1963; Slautterback, 1963), studies of colchicine prompted the identification of tubulin as the subunit comprising these polymeric fibrils (Borisy and Taylor, 1967a; Borisy and Taylor, 1967b). Advances in electron microscopy (EM) and fixation methods subsequently allowed us to postulate how tubulin subunits are arranged within a microtubule (Tilney et al., 1973). Since then, countless studies have given us a detailed picture of the different conformations of tubulin, the structure of growing and shrinking microtubules, and the architecture of the microtubule network in cells.

The quasi-helical symmetry of microtubules allowed for the averaging of tubulin structures within the lattice, facilitating early high resolution EM studies. Advances in EM imaging and analysis, as well as clever tricks such as the decoration of microtubules with kinesin to break the symmetry within the lattice and the use of nucleotide analogs, have allowed scientists to determine the structure of different conformations of tubulin with impressive resolution (see, e.g., (Alushin et al., 2014; Manka and Moores, 2018; LaFrance et al., 2022). Thus, we now know that microtubules are approximately 25 nm wide hollow tubes comprised, in our cells, of 13 laterally-associated protofilaments, each of which is a linear chain of α,β-tubulin heterodimers polymerized head-to-tail to produce filaments with two distinct ends. Both tubulin monomers are structurally similar with a globular body of ∼4 nm diameter that binds GTP and makes contacts with neighbouring tubulin dimers in the microtubule lattice, and a flexible C-terminal tail located at the outside surface of the microtubule. However, only β-tubulin, exposed at the more dynamic plus-ends of microtubules (Mitchison, 1993), hydrolyzes and exchanges its GTP in a hydrolysis cycle intricately coupled to conformational changes in the tubulin dimer (Alushin et al., 2014; Manka and Moores, 2018; LaFrance et al., 2022). These conformational changes facilitate the growth and shrinkage of microtubules; GTP-tubulin incorporates into the microtubule at the growing end and undergoes GTP hydrolysis within the microtubule lattice after some delay, resulting in the presence of a stabilizing cap of GTP-tubulin. If this cap is lost and GDP-tubulin is exposed instead, the microtubule undergoes a catastrophe and rapidly depolymerizes unless it is rescued and re-enters a period of growth. These cycles of polymerization and depolymerization are known as microtubule dynamic instability (Mitchison and Kirschner, 1984).

Preceding structural studies, observations of microtubule dynamic instability are precisely what initially allowed scientists to posit that microtubules are polar; the plus- and minus-ends of microtubules were defined based on their growth speeds; the fast-growing end was termed the plus-end, while the slow-growing end was called the minus-end (Borisy, 1978). While these older studies focused on the distinct biochemical properties of the plus- and minus-end (Allen and Borisy, 1974; Dentler et al., 1974; Bergen and Borisy, 1980; Mitchison, 1993), this polarity arises—importantly—from the head-to-tail association of α,β-tubulin heterodimers within the microtubule and can thus be read along the lengths of these filaments. This allows them to serve as tracks for transport: motor proteins of the kinesin family and dynein can use the asymmetry of these filaments to walk in a directed manner towards either the plus- or minus-end of the filament. Thus, it has long been assumed that the polarity of microtubules is carefully controlled in cells as this would determine the direction of transport. Indeed, we now know that the dynamics and positioning of microtubules can be regulated by a slew of nucleation factors and other microtubule-associated proteins (MAPs) that influence the nucleation, growth, shrinkage, stability, and organization of microtubules. By specifically utilizing different MAPs, cells are able to construct unique architectures of microtubules that are essential for their function. Cycling cells typically have a radial array of microtubules focused at a centrosomal microtubule organizing center (MTOC) or two overlapping asters of microtubules during cell division. Post-mitotic cells such as neurons, on the other hand, tend to have polar arrays of microtubules, likely important to establish and maintain cell polarity and function, at least in part by dictating the transport of cargoes in the cell by kinesins and dynein (Kapitein and Hoogenraad, 2015; Burute and Kapitein, 2019).

Some MAPs, instead of (strictly) controlling microtubule organization or dynamics, regulate kinesin- or dynein-driven transport by inhibiting or activating some of these motor proteins; this forms the basis of the so-called MAP code (Monroy et al., 2020). Importantly, not all motors are similarly affected by these MAPs. For example, while tau and MAP2 have inhibitory effects on both kinesin-1 and kinesin-3 in in vitro reconstitution experiments, MAP7 recruits and activates only kinesin-1, and MAP9 and Septin 9 (SEPT9) activate only kinesin-3 (Karasmanis et al., 2018; Hooikaas et al., 2019; Monroy et al., 2020). Furthermore, doublecortin (DCX) and doublecortin-like kinase 1 (DCLK1) also have inhibitory effects on kinesin-1, but both permit kinesin-3 motility (Monroy et al., 2020). Interestingly, while most MAPs seem to have a limited effect on dynein, MAP9 does inhibit its motility (Monroy et al., 2020). In this way, MAPs localized to different places in the cell can help guide cargoes bound to different motor proteins to their correct destination.

Another aspect potentially controlling microtubule-based transport by kinesins and dynein is the so-called tubulin code or the diversity of tubulin. Our cells encode nine genes for α-tubulin and nine genes for β-tubulin (Janke and Magiera, 2020). These isotypes differ largely in terms of their C-terminal tails, but there is also variability in the M-loop (Roll-Mecak, 2020), a structural component essential for lateral interactions between neighbouring tubulin dimers (Alushin et al., 2014). Of these isotypes, at least five of each are expressed at some point during neuronal development (Hausrat et al., 2020). Moreover, tubulin is subject to a wide array of post-translational modifications (PTMs) that further increase the diversity of the tubulin pool in our cells. Many of these PTMs occur on the C-terminal tails of the tubulin dimer such as the removal of the terminal tyrosine (detyrosination) (Arce et al., 1975; Hallak et al., 1977; Gundersen et al., 1984) or additionally of the adjacent glutamate (∆2) (Paturle-Lafanechere et al., 1991), but some also occur on the body of tubulin. The best-studied example of this is the acetylation of Lysine 40 on α-tubulin (αK40). This residue resides inside the microtubule lumen and its acetylation is thought to be a marker for long-lived, stable microtubules (Schulze et al., 1987). This PTM is well-studied in part because of kinesin-1’s strong preference for acetylated microtubules in cells (Reed et al., 2006; Cai et al., 2009; Tas et al., 2017); however, the motor is indifferent to this modification in vitro (Walter et al., 2012; Kaul et al., 2014). This suggests that kinesin-1 may not be recognizing a specific feature of acetylated microtubules directly, but rather that the microtubules that kinesin-1 prefers are typically also acetylated in cells. Alternatively, some (structural) aspect of acetylated microtubules in cells may be lost in vitro, for example, due to the use of Taxol, a drug that binds to microtubules and suppresses their dynamics, or GMPCPP, a non-hydrolyzable GTP analog, to stabilize microtubules. Together, the tubulin isotypes and their PTMs make up the so-called tubulin code (Janke and Magiera, 2020; Park and Roll-Mecak, 2018; Roll-Mecak, 2020).

Microtubules composed of different tubulin isotypes, bearing a variety of PTMs, and decorated by a slew of MAPs are organized into a spectacularly asymmetric array of microtubules in neurons. The organization of the neuronal microtubule cytoskeleton is thought to be essential for neuronal polarity, a prerequisite for their function. Thus, much scientific effort has focused on understanding precisely how the microtubule cytoskeleton in neurons is organized and in dissecting the processes that allow this network to be built and maintained. In this review, we will highlight important studies on the architecture of the neuronal microtubule cytoskeleton and the technological advancements that have facilitated these discoveries. Specifically, we will focus on the heterogeneity of microtubules and aim to define neuronal microtubule subsets, including their characteristics and functional roles. We will also speculate on how different aspects defining subsets can be coordinated and how distinct subsets can be maintained simultaneously in the cell.

Determining microtubule orientation

In neurons, the differential localization of cargoes to the axon and dendrites is essential for neuronal polarity and likely relies on the directed transport of these cargoes by microtubule-based motor proteins that can enter the axon and/or the dendrites (Lipka et al., 2016). How this transport is directed is a multi-faceted problem, but fundamentally depends on the orientation of microtubules in the neurites: if microtubules are oriented plus-end-out, plus-end-directed motor proteins such as kinesin-1 and kinesin-3 will drive anterograde (away from the cell body) transport, while minus-end-directed motor proteins such as dynein will drive retrograde (towards the cell body) transport; however, if the orientation of the microtubules is reversed, the opposite holds true. Thus, it is important to consider how microtubules are oriented in neurons.

Calculations based on the diffraction patterns from electron micrographs of microtubule doublets provided an early indication that microtubules are polar structures—albeit with some uncertainty (Amos and Klug, 1974). Since then, a variety of techniques have facilitated the visualization of microtubule orientation both in vitro and in cells.

Molecular arrowheads

One technique to determine the orientation of a polar filament is to decorate it with a molecular arrowhead. That is, to decorate the filament with some other component that is asymmetrical on a larger scale that is resolvable with a given imaging method. In the actin field, myosin heads decorating actin filaments fulfilled precisely this role; in electron micrographs, myosin appeared as arrowheads (or chevrons) bound along actin filaments with a uniform orientation, suggesting that the underlying filaments were also polar along their length (Huxley, 1963). The appearance of these decorated actin filaments was also responsible for the nomenclature of the two ends of the filament: “pointed” and “barbed.”

What we now know to be axonemal dynein fragments are similarly able to decorate microtubules with an asymmetric tilt (Murphy and Borisy, 1975) (Figure 1A). While their consistent tilt along the length of microtubules is indicative of microtubule structural polarity, this technique never gained traction in the microtubule field for ascertaining the orientation of microtubules. Years later, however, the kinesin motor domain was used to serve a similar purpose in higher resolution EM studies (Alushin et al., 2014). Until a resolution of about 8 Å, α- and β-tubulin are structurally highly similar, making them difficult to distinguish in reconstructions to break this resolution barrier; however, decorating the microtubule with the kinesin motor domain allows tubulin dimers to be distinguished (i.e., the polarity of the microtubule is known) and allows for the identification of the microtubule seam. Thus, the kinesin motor domain served to break the symmetry of the microtubule and allowed for high resolution microtubule structures to be solved using cryo-EM.

www.frontiersin.org

FIGURE 1. Methods for visualizing the polarity of (neuronal) microtubules. (A) Decoration of microtubules polymerized from porcine brain tissue with dynein heavy chain fragments. Note the uniform tilt of the dynein fragments along the length of a given microtubule, indicating the structural polarity of the filaments. Taken from (Murphy and Borisy, 1975). (B) The moiré patterns from microtubules grown from purified calf brain tubulin and nucleated off of isolated centrosomes such that their polarity is known. All microtubules oriented plus-end-up. Chevrons drawn in to help visualize the chevrons in the moiré patterns. When microtubules have 13 protofilaments they run parallel to the microtubule axis, but when they have another number of protofilaments, they have a supertwist, producing arrowheads of a uniform direction; however, the direction of these arrowheads depends on the number of protofilaments, which dictates the handedness of the supertwist. Taken from (Chrétien et al., 1996). (C) A 3D reconstruction of a 15 protofilament microtubule, as well as a corresponding axial view and a projection of a thin slice from this reconstruction, both viewed from the plus-end towards the minus-end. Note that the tubulin monomers are tilted counter-clockwise when viewed from this end. This tilt is independent of protofilament number. Taken from (Sosa and Chrétien, 1998). (D) Cross-section of an axon from a rat hippocampal neuron with microtubules subject to hook decoration. Neurons are lysed and a hooking buffer containing additional brain tubulin is added such that it forms hooks along the existing microtubules. Note that in this axon viewed from the growth cone towards the cell body, all the hooks are oriented clockwise, indicating that the plus-ends of the microtubules are pointing towards the growth cone; however, many microtubules do not have any hooks, so their orientation is unknown. Taken from (Baas et al., 1988). (E) Diagram, confocal snapshots, and kymograph from a laser severing and EB comet tracking experiment in an axon (proximal end pointing downwards) from a rat hippocampal neuron. In the diagram, stable microtubules have no EB comet preceding laser severing, while following laser severing, a microtubule with an EB comet can grow from the stable template. Position and time of laser severing indicated in blue. The kymograph shows that all comets move anterogradely both before and after laser severing, but the frequency of comets increases substantially after severing. Taken from (Yau et al., 2016). (F) Cross-section of part of a dendrite from a rat hippocampal neuron with microtubules subject to hook decoration as in (D). Note that in this dendrite viewed from the growth cone, some hooks are pointing clockwise, while others are pointing counter-clockwise, indicating that microtubules are of mixed orientation. Again, many microtubules do not have hooks, so their orientation is unknown. Taken from (Baas et al., 1988). (G) Diagram, confocal snapshots, and kymograph from a laser severing and EB comet tracking experiment in a dendrite (proximal end pointing downwards) from a rat hippocampal neuron. In the dendrite, there are EB comets moving both anterograde (green) and retrograde (red), as seen in the snapshots and the kymographs. Again, the comet frequency increases massively after laser severing. Taken from (Yau et al., 2016). (H) motor-PAINT reconstruction from a dendrite in a rat hippocampal neuron. Microtubules reconstructed in green had kinesin motors moving from left (proximal) to right (distal) and are thus plus-end-out, while those coloured red are pointing in the opposite direction. Taken from (Tas et al., 2017). (I) The intensity profiles along the line indicated in (H), showing that minus-end-out microtubules are bundled centrally, while plus-end-out (minus-end-in) microtubules are bundled peripherally. Taken from (Tas et al., 2017). (J) A graph showing the quantification of the percentage of microtubules (based on motor localizations) that are oriented minus-end-out in the proximal, middle, and distal parts of dendrites. This percentage is always close to 50%, suggesting that microtubule orientations are mixed 50–50 along the length of the dendrites. Taken from (Tas et al., 2017).

Analysis of supertwist of protofilaments

Analysis of moiré patterns of microtubules also suggested that microtubules were polar, as the supertwist of protofilaments had a consistent direction along the length of the microtubule, forming arrowheads (Mandelkow et al., 1986) (Figure 1B); however, establishing whether these arrowheads were pointing towards the minus- or plus-end of the microtubule was a different challenge because the handedness of this supertwist depends on the number of protofilaments (which can vary for in vitro polymerized microtubules) and because there is no supertwist if the microtubule has 13 protofilaments (Chrétien et al., 1996; Sosa and Chrétien, 1998).

Hook decoration

Hook decoration, another EM-based technique, was fundamental in determining the orientation of microtubules in neurons. This technique also relies on adding a larger, more easily visible structural asymmetry to microtubules. In vitro work with isolated axonemes had demonstrated that, when soluble tubulin is added to microtubules in a so-called “hooking buffer,” this tubulin polymerizes into hooks along the surface of the microtubules. Importantly, because the fast- and slow-growing ends could be distinguished in these experiments, it became apparent that when viewed from the fast-growing plus-end toward the slow-growing minus-end, the hooks are right-handed (Heidemann et al., 1980; Heidemann and McIntosh, 1980). Soon after, this technique was used to study the orientation of axonal microtubules in different types of neurons, demonstrating that they are uniformly plus-end-out (Burton and Paige, 1981; Heidemann et al., 1981; Baas et al., 1988) (Figure 1D). Unlike axons, dendrites of vertebrate neurons were found to contain microtubules of mixed polarity with approximately half oriented plus-end-in and half oriented plus-end-out (Baas et al., 1988) (Figure 1F).

Analysis of tubulin tilt in microtubule cross-sections

Interestingly, with high-resolution three-dimensional reconstructions of microtubules, the axial view of microtubules (cross-sections perpendicular to the length) also provides insight into the polarity of microtubules, as tubulin subunits are tilted counter-clockwise when looking from the plus-end towards the minus-end, and clockwise when viewed the other way around (Sosa and Chrétien, 1998) (Figure 1C). Unlike the assessment of the arrowheads present in moiré patterns due to the supertwist of protofilaments, this method is independent of the number of protofilaments. It has recently been applied in detailed cryo-electron tomography studies of axons in D. melanogaster and mice neurons, allowing for the assignment of microtubule polarity along with the observation of, for example, their end tapers and associated organelles (Foster et al., 2021).

EB comet tracking

The findings on the orientation of microtubules in axons and dendrites made using hook decoration were corroborated by evidence from another technique used to assess microtubule orientation, this time based on light microscopy: end-binding (EB) comet tracking (Stepanova et al., 2003). The over-expression of fluorescently-tagged plus-end tracking proteins (+TIPs), proteins that selectively track the growing plus-ends of microtubules, allowed for the visualization of the direction of microtubule growth—and hence their orientation—in live cells [and even in mice (Kleele et al., 2014)] using light microscopy. The technical simplicity of this method allowed scientists to ascertain the orientations of microtubules in other types of neurons, including those of D. melanogaster and C. elegans. Here, unlike in vertebrate neurons, dendrites were found to contain uniformly minus-end-out microtubules (Stone et al., 2008; Hill et al., 2012). Interestingly, because observations are made with live cells, one could in theory manipulate the cells (e.g., treat them with a drug affecting the microtubule cytoskeleton) and have a read-out of the effects, if any, on microtubule organization. However, the technique does not produce images of the overall architecture of the microtubule network and only provides reliable information on the orientation of dynamic microtubules.

To bypass this limitation and also acquire orientation information about stable microtubules, EB comet tracking can be combined with laser cutting. To do so, microtubules are first severed with the hope that the newly severed microtubule ends serve as seeds for microtubule growth regardless of whether the severed microtubules were themselves dynamic or stable (Yau et al., 2016) (Figures 1E, G). However, laser severing is quite harsh on cells and might also lead to the release of Calcium ions from the endoplasmic reticulum (ER), which could in turn activate a variety of signalling pathways. Alternatively, visualizing the motility of cargoes such as Rab4-positive structures can also provide insight into how microtubules are connected between dendritic branches and, assuming that, unlike EB comets, these cargoes (also) move on stable microtubules, complement EB studies in assessing microtubule organization (Stone et al., 2008).

Second-harmonic generation

Another technique used to visualize the parallel arrangement of microtubules in neurons that has provided insights into the organization of microtubules in brain slices of mice is second-harmonic generation (SHG) microscopy (Dombeck et al., 2003; Kwan et al., 2008). Here, trains of laser pulses are applied to live neurons or brain slices for imaging. Importantly, this technique is label-free and signal is generated via the structural polarity of the microtubules themselves. The signal from two adjacent parallel microtubules interferes constructively to reinforce the signal, whereas the signal from two adjacent antiparallel microtubules interferes destructively to minimize the signal. It is important to note, however, that both antiparallel and randomly mixed arrays would generate no or low SHG signal, and both 100% plus-end-out and 100% minus-end-out microtubule arrays would also generate a similar signal, restricting the use of this technique. Moreover, to interpret the signal, knowledge about the spatial distribution of microtubules (e.g., from EM studies) is required, and to put findings in context, the expression of a fluorescent marker or the subsequent staining for markers such as tau or MAP2 is also needed.

The strengths of SHG microscopy are that it works (only) with live samples and can be used to image in thick samples (i.e., brain slices). Thus, it has been used to image brain slices of mice up to 18 months old (Kwan et al., 2008). This revealed a trend in which some dendritic microtubules—specifically, in CA1 apical dendrites and the layer V cortex—become increasingly polarized within the first ∼4 months of a mouse’s life and then remained stably ∼80% polarized in older mice. In contrast, this was not observed in cultures of hippocampal neurons, in agreement with the work described above, indicating the importance of studies in brain slices and suggesting that later in life, vertebrate neurons might more closely resemble those of D. melanogaster or C. elegans with their uniform minus-end-out dendritic microtubule arrays.

motor-PAINT

Most recently, motor-based Point Accumulation for Imaging in Nanoscale Topography (motor-PAINT), a technique simultaneously allowing for the super-resolution reconstruction of microtubules and the determination of their orientation, was developed (Tas et al., 2017). In this technique, the membranes of cells are permeabilized and the cells are gently fixed such that the structure of the microtubule cytoskeleton is preserved. Subsequently, purified fluorescent kinesin motors are allowed to explore the microtubule network and imaged for many (10,000–20,000) frames. In each frame, the particles are localized with sub-pixel precision, allowing for a super-resolution reconstruction of the microtubule network. By additionally linking the localizations of the motors in subsequent frames, the orientations of the underlying microtubules can be inferred from the directions of the tracks.

As this technique uses light microscopy rather than EM, it is easier to acquire large fields of view along the lengths of neurites and, unlike EB comet tracking, it provides information about the orientation of all microtubules regardless of dynamicity. Furthermore, the super-resolution images produced by motor-PAINT and its higher throughput nature provided insight into the ultra-structural organization of microtubules in neurites. Thus, this technique made it evident that in dendrites, microtubules are organized such that plus-end-out microtubules form bundles peripherally, while plus-end-in microtubules are bundled centrally (Tas et al., 2017) (Figures 1H, I). Moreover, this holds true over sections along the length of the dendrite (Figure 1J). One disadvantage of motor-PAINT, however, is that it requires the microtubules to be only gently fixed and accessible for kinesin binding, making it difficult to combine with other labelling or imaging strategies that would allow for the simultaneous localization of different MAPs or PTMs. However, this can be overcome by using post-fixation and re-imaging the same cell with a different super-resolution technique (Tas et al., 2017). In its current form, motor-PAINT also relies on total internal reflection fluorescence (TIRF) microscopy to minimize the effect of the high background signal due to diffusing motors, making it difficult to image throughout the cell volume or, for example, in brain slices.

What the bundled, locally-uniform organization of microtubules described above cannot explain, however, is why kinesin-1 can only drive transport into the axon, while kinesin-3 non-discriminately enters the axon and the dendrites (Lipka et al., 2016 EMBO J). To better understand this, we must also consider the diversity of microtubules present in cells, including the collection of tubulin isotypes, PTMs, and MAPs that they are associated with.

High resolution studies of microtubule organization and diversity

In addition to their orientation, controlling microtubule architecture in cells also involves controlling the distribution of tubulin isotypes and of microtubules bearing different PTMs or MAPs. These aspects generate microtubule diversity, allowing microtubules to be specialized to fulfill different functions in cells. Sub-cellular control over the localization or enrichment of different MAPs and PTMs has been recognized for many years and is thought to be important in establishing neuronal polarity.

Early studies using diffraction-limited light microscopy revealed the asymmetric distributions of MAP2 and tau to dendrites and axons, respectively (Binder et al., 1986; Matus et al., 1986). Diffraction-limited microscopy can thus reveal large scale distinctions in the distribution of a MAP or PTM, such as its localization to a given neurite, whether it is enriched proximally or distally along a neurite, or whether it is, for example, enriched near branch points. Early work from cells with radial microtubule arrays, however, already made it apparent that adjacent microtubules can “look” very different, with studies indicating that microtubules are not uniformly decorated with PTMs, but that these modifications tend to segregate to specific microtubules, as seen with acetylation (Thompson et al., 1984) and detyrosination (Gundersen et al., 1984). However, getting a better understanding of the role and distribution of different types of microtubules in neurons with their densely packed axonal and dendritic microtubule arrays [spacing of ∼20–70 nm (Chen et al., 1992)], requires microscopy techniques with better resolution.

This is an interesting question to consider because neurons critically depend on microtubule-based transport, and the kinesin and dynein motors driving this transport have been shown to preferentially use microtubules decorated by a specific PTM and be activated or inhibited by the presence of different MAPs. Thus, the tubulin code and MAP code are likely important regulators of microtubule-based transport in neurons; however, the precise roles of many MAPs and PTMs remain largely unresolved.

Single-molecule localization microscopy

Understanding the track selectivity of motor proteins and how PTMs and MAPs regulate neuronal transport requires us to visualize how different types of microtubules are organized in these cells. To resolve the precise organization of microtubules, super-resolution techniques are needed, especially when studying bundled or dense microtubule arrays, such as are found in neuronal processes. Multiple studies have used PAINT or Stochastic Optical Reconstruction Microscopy (STORM) approaches to achieve this. In general, these techniques do not allow for individual microtubules within bundles to be resolved as the linkage error resulting from the physical size of the primary and secondary antibodies increases the effective microtubule diameter by ∼25 nm, such that the closely bundled microtubules in dendrites are blurred together (Figure 2A). One way to overcome this is to use labelled nanobodies, which can reduce the linkage error from ∼12.5 to ∼2.5 nm (Mikhaylova et al., 2015) (Figure 2A). Thus, one could imagine generating multiple nanobodies to simultaneously resolve different types of PTMs or MAPs in neurons.

www.frontiersin.org

FIGURE 2. Methods for visualizing aspects of the MAP code and the tubulin code. (A) An illustration showing different labelling strategies [primary + labelled secondary, labelled primary, and labelled nanobody (here VHH)] used in single molecule localization microscopy approaches such as STORM to highlight their effect on the effective microtubule diameter. Experimental set-up to create microtubules cross-linked with a known distance of 30–40 nm between microtubules or 55–65 nm between microtubule centers using MAP65/Ase1 from A. thaliana. An experimental comparison of the observed peak-to-peak distances for microtubules crosslinked with MAP65 when imaged using a labelled primary antibody (left) or a labelled nanobody (right). Note that microtubules can be more easily distinguished with the nanobody. Taken from (Mikhaylova et al., 2015). (B) Volumetric 3D rendering of microtubules in a dendrite as imaged with four-fold expansion microscopy, as well as corresponding cross-sections. Shown are total tubulin, tyrosinated tubulin, acetylated tubulin, and an overlay. Note that acetylated microtubules are enriched centrally, while tyrosinated microtubules are enriched peripherally. Taken from (Katrukha et al., 2021). (C) A Stimulated Emission Depletion (STED) microscope image of a dendrite from a DIV7 rat hippocampal neuron similarly showing the central and peripheral enrichment of acetylated and tyrosinated microtubules, respectively. This is also visualized in the line scan below. Note, however, that the resolution here is much worse than with expansion microscopy. Taken from (Tas et al., 2017). (D) STED image of a DIV9 rat hippocampal neuron stained for Δ2, tyrosinated, and acetylated tubulin. Interestingly, Δ2 tubulin is also enriched centrally in dendrites, perhaps more tightly than acetylated tubulin. Taken from (Katrukha et al., 2021). (E) A correlative experiment in which kinesin-1 (KHC [1–560]-3xmCitrine) motility was imaged in live cells (observed tracks shown as a standard deviation map) and cells were subsequently fixed and stained for total tubulin and acetylated tubulin. The microtubules utilized by kinesin-1 are predominantly acetylated. Taken from (Cai et al., 2009). (F) The footprint of the microtubule-binding domain of tau and the structure of MAP7 as solved by cryo-EM. Tau binds at the very ridge of the protofilament, whereas MAP7 binds away from this site towards the lateral contacts. Depending on the isoform, tau has either three or four microtubule-binding pseudo-repeats that bind along a single protofilament, only one of which is shown. As each microtubule-binding domain binds across an inter-dimer interface, adjacent tubulin dimers can be stapled together by the microtubule-binding repeats. The microtubule binding region of MAP7 similarly spans across an inter-dimer and an intra-dimer interface, but the protein does not stretch as far along a protofilament. Taken from (Kellogg et al., 2018) and (Ferro et al., 2022). (G) Cryo-electron tomography of in vitro microtubules polymerized from purified tubulin in the presence of GMPCPP and MAP6, showing that MAP6 forms particles (indicated by arrowheads) at regularly spaced intervals inside the microtubule lumen (also shown in the cross-section below). Taken from (Cuveillier et al., 2020). (H) Maps of ∼96% acetylated and < 1% acetylated microtubules as solved by cryo-EM. Note that the overall structures of the tubulin dimer look very similar regardless of acetylation status. However, changes can be observed in the loop in the lumen of the microtubule on which αK40 is found in that the loop becomes more ordered upon acetylation. This can be seen here in that more residues appear in the reconstruction of the acetylated tubulin dimer, thereby suggesting that they are less flexible after acetylation. Taken from (Eshun-Wilson et al., 2019).

Individual microtubules have also been distinguished in bundles using an approach similar to motor-PAINT (Balabanian et al., 2017), and motor-PAINT has recently been combined with MINimal photon FLUXes (MINFLUX) microscopy (see section “Live-cell markers and microscopy” below) as the improved resolution of this technique allows microtubules to be distinguished more easily (Deguchi et al., 2023). While this method does not allow different types of microtubules (e.g., those decorated by different PTMs) to be distinguished directly (as kinesin-1 selectivity is not preserved in vitro), it allows single microtubules and bundles to be distinguished and can be combined with subsequent antibody stainings and STORM to visualize different PTMs. This can provide novel insights, such as the preferential bundling of acetylated microtubules in COS-7 cells (Balabanian et al., 2017), and when looking at microtubule orientation, that plus-end-in microtubules, which are bundled centrally in dendrites, are predominantly acetylated, while peripheral plus-end-out microtubules are enriched in tyrosinated tubulin (Tas et al., 2017). These correlative findings suggests that, while microtubules in dendrites might be of mixed orientation, local uniformity in microtubule type and orientation through bundling could create “super-highways” for microtubule-based transport.

Expansion microscopy

Another increasingly popular method that greatly improves the effective resolution in light microscopy is expansion microscopy (ExM) (Chen et al., 2015; Tillberg et al., 2016). Recent improvements in the method have allowed for up to ten-fold expansion (Damstra et al., 2022; Li et al., 2022; Klimas et al., 2023) and near single-microtubule resolution in dendrites even with four-fold expansion using Flip-ExM, facilitating a precise mapping of how acetylated and tyrosinated microtubules are organized in dendrites, including quantifying the absolute numbers of microtubules (Jurriens et al., 2021; Katrukha et al., 2021) (Figure 2B). As seen previously with STED (Figure 2C) and STORM (Tas et al., 2017), this study confirmed with improved (3D) resolution that in dendrites, acetylated microtubules are bundled centrally, Δ2 tubulin is enriched centrally (Figure 2D), and tyrosinated microtubules are located peripherally. Moreover, this work found that most microtubules are accounted for when considering acetylated and tyrosinated microtubules (Katrukha et al., 2021), and because other PTMs are also found in dendrites, this—not surprisingly—indicates that microtubules must bear multiple PTMs. Thus, while acetylation is often considered a marker for stable microtubules, perhaps acetylated and Δ2-modified microtubules are even more stable. It is unclear, however, how many acetylated microtubules are Δ2-modified and vice versa. Resolving single microtubules will be important to categorize them into different groups based on one or more properties.

One of the key advantages of ExM over many other super-resolution techniques is that the effective resolution is improved not only in x and y, but also in z. Thus, it should be possible to determine, for example, whether MAPs preferentially localize to a specific subset of microtubules, are enriched on one side of a microtubule or if they are located inside a microtubule rather than on its surface. Protofilament-specific MAP and microtubule inner protein (MIP) localization has been seen on ciliary doublet microtubules using cryo-EM (Ma et al., 2019), and it will be interesting to see if this protofilament-specific localization of MAPs and/or MIPs can also be seen for neuronal microtubules to define highways for different types of cargoes. Getting such precise information with ExM may require, for example, the use of nanobodies and/or post-expansion labelling to reduce the linkage error, which is also expanded (Tillberg et al., 2016; Zwettler et al., 2020). Thus, with typical primary and secondary antibody stainings using Ten-fold robust expansion (TREx) microscopy (Damstra et al., 2022), the linkage error of 12.5 nm on either side of the microtubule (Mikhaylova et al., 2015) is expanded to about 125 nm, so the microtubule’s observed diameter is double its true diameter. When using nanobodies, which result in a smaller linkage error of about 2.5 nm (Mikhaylova et al., 2015), the linkage error would be 6x less. If post-expansion labelling is used, the linkage error would not be expanded, decreasing its contribution to the observed microtubule diameter ten-fold such that the linkage error would contribute < 10% of the observed microtubule diameter.

Cryo-electron microscopy

Determining precisely how MAPs bind the microtubule or observing structural changes in the tubulin dimer, however, requires the structural context and resolution provided by EM. Improvements in the technique and the analysis methods have allowed scientists to solve the structure of tubulin within the microtubule under various conditions, as well as the structure of MAPs decorating the microtubule lattice, with impressive detail. For example, the structure of Protein Regulator of Cytokinesis 1 (PRC1), tau, and MAP7 bound to the microtubule have been solved with near atomic resolution, with PRC1 binding at the intradimer interface, tau binding along the ridge of a protofilament and linking multiple subsequent tubulin dimers, and MAP7 binding along a protofilament, but extending less far along it and binding more towards the inter-protofilament contact sites rather than at the ridge (Kellogg et al., 2016; Kellogg et al., 2018; Ferro et al., 2022) (Figure 2F). This detailed structural information provides insight into how these proteins function. For example, one can precisely compare the binding footprint of these proteins with that of kinesin or dynein to better understand how these proteins might interact on the microtubule surface. Such analysis has been done for MAP4, suggesting that it can bind microtubules without impeding the binding of kinesin-1 (Shigematsu et al., 2018). In another study, a combination of EM techniques was used to study MAP6 function, revealing that this MAP is actually a MIP as it localizes to the inside of microtubules both in vitro and in neurons, that it can coil microtubules, and that it introduces apertures along the microtubule lattice (Cuveillier et al., 2020) (Figure 2G).

In addition to revealing these precise interactions, EM can also provide information about the microtubule lattice under different conditions. It is known that tubulin can adopt both an expanded (∼8.4 nm) and a compacted (∼8.2 nm) form at least in part depending on its bound nucleotide, with GTP-bound tubulin being more expanded than GDP-bound tubulin (see also “Defining stable microtubules” below) (Alushin et al., 2014; LaFrance et al., 2022). Interestingly, however, it appears that certain MAPs, such as the neuronal MAPs tau and MAP2, can induce local compaction of the lattice at points where they bind cooperatively to form so-called islands or envelopes (Siahaan et al., 2022). In contrast, kinesin-1 has been shown to expand the microtubule lattice, at least in vitro, with ∼10% lattice occupancy (Peet et al., 2018; Shima et al., 2018). Furthermore, a kinesin-1 rigor (with a mutation that prevents it from hydrolyzing ATP, rendering it non-motile) has been shown to localize preferentially to microtubules with an expanded lattice in cells (de Jager et al., 2022). Thus, there is some flexibility in the structure of tubulin and this could serve as a means for proteins to bind cooperatively and communicate allosterically via the microtubule lattice (Brouhard and Rice, 2018). Indeed, studying how these proteins interact to induce lattice compaction or expansion has been the focus of some recent work (Siahaan et al., 2022) and it will be interesting to see if this differs, for example, between dendrites and axons, near synapses or branch points, and on different microtubule subsets.

In theory, PTMs could also impact tubulin structure; however, most occur on the flexible C-terminal tails of α- or β-tubulin, making it less obvious how local structural changes, if any, could be propagated to the rest of the tubulin dimer. Despite this, the C-terminal tails themselves adopt a variety of transient conformations and this “structural landscape” could be altered by PTMs such as polyglutamylation. This could in turn influence how proteins interact with the microtubule (Bigman and Levy, 2020; Chen et al., 2021). One commonly-studied PTM that occurs on the body of the tubulin dimer, albeit in the lumen of the microtubule, is the acetylation of αK40 and the effects of this PTM on the structure of tubulin have been investigated; however, it appears that the acetylation of this residue does not lead to large-scale allosteric structural changes in the tubulin dimer that could be easily recognized by proteins binding to the outside of the microtubule. Instead, it leads to more subtle changes in the lateral interactions between tubulin dimers (Howes et al., 2013; Eshun-Wilson et al., 2019) (Figure 2H), which could still be detected, e.g., by some MIPs or MAPs that bind between protofilaments.

Correlative live and fixed cell approaches

Despite these advances in imaging methods and insights into how MAPs and PTMs influence microtubule structure, comprehensive studies allowing us to query the precise sub-cellular distribution of these features as was done for the distribution of acetylated and tyrosinated microtubules in dendrites (Katrukha et al., 2021) or understand the functional significance of their structure and localization have been limited. This is due in part to the fact that most of the techniques discussed above do not allow for live-cell imaging, which would be beneficial to assess microtubule-related functions (e.g., cargo motility). For example, it is unclear whether different dendritic cargoes preferentially use either the central highways of plus-end-in acetylated microtubules or the peripheral highways of plus-end-out tyrosinated microtubules.

To bypass this, some correlative work has been done in which transport is observed in live cells, which are subsequently fixed and the same cell is imaged by a super-resolution technique such as STORM to assess how cargoes traverse microtubule crossings (Bálint et al., 2013) or to better trace microtubules along their lengths and assess certain characteristics such as damage sites (Hao et al., 2020). Correlative studies with diffraction limited light microscopy were also done to demonstrate that kinesin-1 preferentially moves on acetylated microtubules in vivo (Reed et al., 2006; Cai et al., 2009) (Figure 2E). Furthermore, such correlative studies also revealed that lysosomes are enriched on detyrosinated microtubules, and by knocking down motors of the kinesin-1 (KIF5B), kinesin-2 (KIF3A), and kinesin-3 (KIF1B) family, it was established that KIF5B is responsible for this enrichment on detyrosinated microtubules (Mohan et al., 2018). This suggests that this motor preferentially moves on detyrosinated microtubules [which are also usually acetylated (Katrukha et al., 2021)] even when bound to endogenous cargo that is simultaneously bound to other motors. However, most of this work has been done in cell culture lines with a radial microtubule array rather than in neurons, as the density of the microtubule cytoskeleton in neurons makes it difficult to establish precisely which microtubule a cargo was moving on.

Live-cell markers and microscopy

Because the microtubule cytoskeleton is constantly rearranging, it would be beneficial to develop markers to visualize microtubule diversity in live cells. To this end, CRISPR-based approaches (Jinek et al., 2012; Doudna and Charpentier, 2014; Willems et al., 2020; Droogers et al., 2022) will be important to determine the endogenous distribution of MAPs and tubulin isotypes, but this cannot be used to visualize PTMs or whether microtubules are in an expanded or compacted state. One approach to generate a live cell marker for a given PTM is to screen a library of mutants for a binder that specifically associates with, e.g., tyrosinated microtubules and then express a fluorescently-tagged version thereof in cells (Cassimeris et al., 2013; Kesarwani et al., 2020). One could also imagine a more targeted approach by engineering fluorescent probes based on proteins or protein domains that have been observed to specifically associate with given PTMs, e.g., the CAP-Gly domain which has been shown to associate with tyrosinated tubulin (Peris et al., 2006) or, for example, engineering mutant enzymes [e.g., αTAT1 or a Tubulin Tyrosine Ligase Like (TTLL) family member] that specifically associate with deacetylated or non-glutamylated tubulin but do not themselves alter the PTM status of tubulin. These live cell markers, however, would have limited utility in neurons when trying to investigate the precise partitioning of these modifications within neurites unless combined with super-resolution imaging techniques that allow for near single-microtubule resolution. Most of these techniques (e.g., STORM and PAINT) are incompatible with live cell imaging because they often require special buffers or high laser powers that are damaging to cells and/or have poor time resolution. One technique that is promising for live-cell super-resolution microscopy is MINFLUX (Gwosch et al., 2020; Schmidt et al., 2021). Here, an oscillating donut-shaped excitation beam and a confocal pinhole detector are used to calculate the position of a given dye molecule with ∼1–3 nm resolution. Currently, however, this technique is limited to single wavelengths, small fields of view, and is slow because it iteratively localizes or tracks each dye molecule in turn. Despite this, it has been recently used to visualize the stepping of kinesin-1 both in vitro (Wolff et al., 2022) and in vivo (Deguchi et al., 2023), demonstrating its promise for visualizing dynamics with incredible resolution.

Studying microtubule dynamics and stability

These and other live-cell imaging techniques will also aid the study of another hallmark property of microtubules: their dynamic instability. This ability to rapidly transition between growth and shrinkage allows them to quickly reorganize into drastically different architectures, as is beautifully exemplified with the mitotic spindle or during neurite outgrowth or axonal pruning. However, microtubules do not all have the same dynamic behaviour. Early on, the micro-injection of tubulin into fibroblast cells allowed scientists to assess the turnover rate of microtubules, revealing that most microtubules have a half-life of ∼10 min, while a smaller sub-population has a half-life of ∼1 h (Schulze and Kirschner, 1987). Thus, the existence of a stable subset of microtubules was identified in some cell lines. Furthermore, it seemed as though these were two distinct subsets given that the micro-injected tubulin either fully labelled microtubules along their lengths or was completely absent from other microtubules. This was also the case after some hours, although the number of unlabelled microtubules decreased with time, suggesting that they do turnover at longer timescales (Schulze and Kirschner, 1986; Schulze and Kirschner, 1987). These stable microtubules were further described to be clustered around the center of the cell and curly in comparison to the rather straight dynamic microtubules (Schulze and Kirschner, 1987). Despite the early discovery of these stable microtubules, two matters complicated their study, especially in neurons: first, there is also evidence that these are not two distinct populations of microtubules, but rather that many microtubules have a stable base and a dynamic plus-end (Baas and Black, 1990; Baas and Ahmad, 1992; Ahmad et al., 1993; Baas et al., 2016; Qiang et al., 2018); and second, the tools to study the behaviour of this rather small subset of microtubules, particularly in live cells, have been limited. This is because these stable microtubules are largely found in areas where the microtubule density is highest (e.g., near the MTOC in cells with a radial microtubule array and in axons and dendrites); because their slow turnover rate makes them difficult to label via pulse-chase type methods; and because they are massively outnumbered by dynamic microtubules in most cell types studied, making them difficult to visualize.

Studying microtubule dynamics in vivo

Studying the dynamics of dynamic microtubules, however, has been successful both in vitro and in vivo. To study their polymerization in vivo, EB comet tracking can be used. Here, the overexpression of EB1 or EB3 at low levels allows one to visualize the growing plus-ends of microtubules by the selective association of EBs immediately behind the GTP cap of growing microtubules (Stepanova et al., 2003; LaFrance et al., 2022) (Figures 1E, G). Tracking these EB comets allows you to assess the growth rates of microtubules, the frequency of growth events, the orientation of dynamic microtubules, and even estimate more detailed parameters such as the rate of GTP hydrolysis based on the comet decay length (Duellberg et al., 2016; Roostalu et al., 2020). EBs, however, also influence microtubule dynamics by promoting microtubule growth, while simultaneously increasing the catastrophe frequency (Maurer et al., 2012; Maurer et al., 2014). One way to minimize this effect is to express an SxIP-motif-containing protein, which then localizes to the plus-ends by binding to EB1, rather than expressing EBs directly (Yau et al., 2016). This construct may in turn, however, compete with endogenous EB binding partners.

Studying microtubule dynamics in vitro

To more precisely study microtubule dynamics and how they are influenced, for example, by different MAPs, in vitro assays in which templated microtubule growth is observed using TIRF microscopy are commonly performed. Here, short, stabilized microtubule segments (“microtubule seeds”) are attached to a coverslip via an antibody or streptavidin within a flow chamber. Soluble tubulin is then introduced, which, in the presence of GTP, can polymerize onto the ends of these microtubule seeds, producing microtubules that undergo cycles of growth and shrinkage (Gell et al., 2010) (Figure 3A). These assays have given rise to our stereotypical picture of microtubule dynamic instability and have allowed us to understand how a wide variety of proteins or other factors influence microtubule dynamics and vice versa. For example, DCX tip tracks at lower concentrations and in this way helps to ensure that microtubules have 13 protofilaments (Bechstedt and Brouhard, 2012). Such in vitro assays have also contributed to our understanding of kinesin-3 track selectivity, as this motor was demonstrated to preferentially dissociate near the plus-ends of microtubules regardless of initial binding site, and to rarely start runs in this region (Guedes-Dias et al., 2019). This suggests that kinesin-3 prefers to bind to and move along the GDP-tubulin lattices found along the microtubule shaft rather than the GTP-tubulin found in the GTP cap. Because microtubule plus-ends are enriched at en passant boutons in rat hippocampal neurons, this was suggested to improve cargo delivery at these pre-synaptic sites (Guedes-Dias et al., 2019).

www.frontiersin.org

FIGURE 3. Methods for studying the properties of stable microtubules. (A) Schematic showing an in vitro assay to study microtubule dynamics and how it is affected by different MAPs. A microtubule seed stabilized with GMPCPP is immobilized via antibodies/streptavidin on an otherwise passivated coverslip and soluble tubulin (and in this case DCX) are added with GTP, such that the tubulin can undergo cycles of nucleated growth and shrinkage. Below are kymographs of two microtubules with the microtubule seed visible in red, and DCX visible in green at two different concentrations: 10 nM (left) and 100 nM (right). Interestingly, DCX shows different behaviours at these two concentrations. Taken from (Bechstedt and Brouhard, 2012). (B) Image showing tyrosinated and acetylated tubulin stainings in DIV9 rat hippocampal neurons in control conditions (left) or after a 2.5 h treatment with 4 µM nocodazole. Note how without the nocodazole treatment, the overall levels of the tyrosinated and acetylated microtubules is similar, but the low dose nocodazole treatment preferentially depolymerizes dynamic (tyrosinated) microtubules such that very little tyrosinated microtubule signal is present on the right, while the level of acetylated microtubules is less affected. Taken from (Tas et al., 2017). (C) Cryo-EM maps of tubulin within the microtubule lattice (viewed from the lumen) with tubulin bound to GMPCPP (a GTP analog) (left), GDP (middle), or GDP + Taxol (right). Note that overall, the GMPCPP-bound tubulin and Taxol-stabilized tubulin look similar, having a longer lattice spacing (expanded conformation), while GDP-bound tubulin is more compacted. As indicated, both GMPCPP- and Taxol-bound microtubules are stable, while GDP-bound microtubules are not, suggesting that lattice expansion and microtubule stability are intricately connected. Taken from (Alushin et al., 2014). (D) A stage 3 (axon specified) rat hippocampal neuron expressing mRFP-fill (magenta) and photoactivatable-tubulin (green) with converted regions also indicated. The motion of the photoactivated tubulin can be clearly tracked for ∼10 min, as shown in the kymographs underneath for neurites in stage 2, 3, and 4 neurons. This reveals the predominantly retrograde motion of microtubules in neurites except for the one designated as an axon. Taken from (Burute et al., 2022). (E) Experimental set-up and image of mCherry-fill in a stage 2-to-3 mouse hippocampal neuron expressing mCherry-fill and mNeonGreen-CAMSAP3. CAMSAP3-labelled microtubules could be tracked for much longer periods of time (10–70 h) than photoactivated-tubulin, as evidenced by the kymographs showing the faster retrograde flow in all neurites except the axon once specified. Taken from (Schelski and Bradke, 2022). (F) U2OS cells expressing KIF5B-rigor tagged with 2xmNeonGreen fixed and stained for acetylated and total tubulin. Note how the KIF5B-rigor preferentially decorates the subset of microtubules that are acetylated, a commonly used marker for stable microtubules. Taken from (Jansen et al., 2023).

Defining stable microtubules based on their lifetime and structure

If these dynamic microtubules are the stereotypical microtubules that we picture when we think about microtubules, then what exactly are stable microtubules and what makes them stable? We will here use the working definition that stable microtubules are just that: stable. This should be the case when cells are exposed to cold temperatures or treated with low doses of nocodazole (a drug that binds to tubulin and interferes with microtubule polymerization), as has been observed for a subset of microtubules in different types of cells including neurons (Figure 3B), where they are, in dendrites, mostly oriented plus-end-in (Schulze et al., 1987; Tas et al., 2017). Treatment with higher concentrations of calcium ions can also be used to destabilize microtubules, but not all cold-stable microtubules appear to be resistant to calcium treatment—at least in the optic nerve of three-month-old rats (Song et al., 2013). This suggests that there might be different degrees of microtubule stability.

When we think of stability in terms of microtubules, we often think of the stable GTP cap. Based on work with GTP analogs and hydrolysis-deficient mutants (Alushin et al., 2014; LaFrance et al., 2022), it has been suggested that tubulin dimers bound to GTP have an expanded lattice compared to GDP-tubulin (dimer rise of ∼8.4 nm instead of ∼8.2 nm). Similarly, Taxol-stabilized microtubules also resemble GTP microtubules in that they are expanded, albeit in a slightly different manner (Alushin et al., 2014) (Figure 3C). In vitro, microtubules can also be stabilized by expanding their lattice using Taxol or the slowly hydrolysable GTP analog GMPCPP. Because no additional factors are required, this suggests that lattice expansion is directly responsible for stabilizing the microtubule lattice. Earlier w

留言 (0)

沒有登入
gif