Microtubule plus end dynamics – do we know how microtubules grow?

Summary:

Microtubules form a highly dynamic filament network in all eukaryotic cells. Individual microtubules grow by tubulin dimer subunit addition and frequently switch between phases of growth and shortening. These unique dynamics are powered by GTP hydrolysis and drive microtubule network remodeling, which is central to eukaryotic cell biology and morphogenesis. Yet, our knowledge of the molecular events at growing microtubule ends is incomplete. Here we focus on how microtubules grow. We integrate recent ultrastructural and cell biological data to propose a realistic model of growing microtubule ends comprised of structurally distinct but biochemically overlapping zones. We discuss how regulatory proteins recognize different tubulin conformations along the microtubule lattice. Finally, we hypothesize that independent control of the major structural transitions at growing microtubule ends – tubulin dimer addition and subsequent closure of the MT wall – are optimized in cells to achieve rapid physiological microtubule growth.

1. Introduction

Microtubules (MTs) are one of the three major cytoskeleton systems traversing the cytoplasm of all eukaryotic cells. MTs are intrinsically polarized and form a network of directional tracks essential for organizing intracellular motions ranging from long-range organelle transport to chromosome segregation in mitosis. The typical text book drawing shows MT minus ends near the cell center, while plus ends extend outward into the cell periphery. This stereotypic organization is characteristic for many tissue culture cells, but it is less clear how the MT network is organized in complex differentiated cells especially in an intact tissue context.^[1] As might be expected for a central organizer of cell function, the life cycle of the MT network inside cells is complex and governed by many diverse processes. Because the initial step of starting a MT requires the unlikely meeting of several tubulin subunits in the right orientation, cells nucleate MTs through locally activated γ-tubulin templates at centrosomes, other organelles or even the lateral surface of existing MTs.^[2] MT minus ends can remain anchored or be released and moved to other MT-organizing centers.^[3] Unprotected minus ends depolymerize, but can become stabilized by CAMSAP/Patronin proteins.^[4] Cooperation of such minus end stabilizers with enzymes that cut existing MTs can thus amplify the number of MT ends and drive MT network reorganization.^[5] At the other, the plus end, MTs elongate by addition of tubulin subunits. MT plus ends frequently switch between phases of growth and shortening, but in cells MT growth often pauses for fairly long periods of time. Owing to their tubular geometry, MTs are the stiffest intracellular polymer with a rigidity similar to hard plastics.^[6] MTs are under mechanical load in cells and bend and buckle under force, and thus can counterbalance cellular contractile forces. This often overlooked role of MTs as compression elements in cytoskeleton tensegrity has gained new attention by observations of MT behavior in cells embedded in soft extracellular environments^[7–9] and a spring-like action of MTs may contribute to muscle cell relaxation.^[10] In addition, MT plus end dynamics themselves can exert pushing and pulling forces.^[11] Lastly, there are many different tubulin isoforms and assembled MTs age and accumulate a zoo of diverse post-translational modifications that modulate both mechanical and biochemical properties of individual MTs.^[12] Decrypting how all these processes acting on the properties of the MT network are meaningfully integrated in a cell represents an ongoing challenge. However, ultimately the unique dynamic properties of MT plus end elongation lie at the heart of MT network organization and function, and it is surprising how little we know for certain about the complex structural and molecular events that occur at growing MT ends. Summarizing recent biochemical, structural and cell biological data, we propose a molecular scale model of what a growing MT end might look like and hypothesize that independent conformational transitions during MT growth provide cellular control points to shape MT network organization.

2. One end to rule them all: Microtubule plus end dynamics

MTs assemble from α/β-tubulin dimer subunits that are arranged head-to-tail along protofilaments. Lateral interactions of typically thirteen protofilaments form the wall of the hollow MT tube, and decades of work have generated an advanced model of MT structure (Fig. 1). While 13-protofilament MTs in which individual protofilaments are straight and parallel to the central MT axis are the most common, a surprising variety of alternative MT architectures exists in specialized cell types and other species.^[13] Longitudinal MT asymmetry arises from the directional arrangement of α/β-tubulin subunits within the MT wall with the less dynamic minus end terminating in α-tubulin, while β-tubulin is exposed at the more dynamic plus end.

Figure 1:

Basic geometry of a 13-protofilament GDP-MT, the most common MT lattice in metazoan cells. MTs are cylindrical, hollow, and polarized polymers of α/β-tubulin dimers arranged head-to-tail along protofilaments. β-tubulin is exposed at the MT plus end, and α-tubulin at the minus end. Longitudinal interactions occur along, and lateral interactions between protofilaments. In a 13-protofilament MT, protofilaments run parallel to the MT axis, which is potentially essential for long-range transport by MT motor proteins. Because of the invariable lateral offset of neighboring tubulin subunits, this is not the case in ‘supertwisted’ MTs with other protofilament numbers. The helical arrangement of tubulin monomers with a 12 nm pitch results in at least one lattice discontinuity at which heterotypic lateral interactions occur between neighboring α and β-tubulin monomers. This MT seam is shown here between protofilaments 1 and 13. Measurements on the figure are rounded to the nearest nanometer.

2.1. Dynamic instability drives microtubule network organization

Reconstitution of MT polymerization in vitro requires only tubulin, guanosine triphosphate (GTP), and either nucleation templates or sufficiently high tubulin concentrations to overcome the thermodynamic barrier of starting new MTs.^[2] Analysis of the MT length distribution in such experiments with highly purified tubulin first revealed that polymerizing (growing) and depolymerizing (shortening) MTs co-exist at the same tubulin concentration.^[14] This non-equilibrium behavior in which individual MTs undergo cycles of growth and shortening at their plus ends is referred to as MT dynamic instability. With the advent of time-lapse video microscopy, MT dynamic instability was soon also directly observed in living cells.^[15,16] Cellular MT functions are intricately linked to this remarkable dynamicity, which is epitomized by the minute time scale during which cells can reorganize the distributed interphase MT network into a compact mitotic spindle.

The now classic ‘search-and-capture’ hypothesis first formalized the idea that dynamic MT plus ends explore the cytoplasm and interact and become stabilized near specific locations such as the cell cortex in interphase or kinetochores on mitotic chromosomes.^[17] Although ‘search-and-capture’ by itself does not fully explain MT network dynamics, MT plus end growth remains central to cellular MT network organization and adaptability. A complex network of +TIP proteins that specifically accumulate at growing MT plus ends is thought to support these interactions. For example, a platform of cell-matrix adhesion-associated proteins interacts with MT +TIP complexes and captures and stabilizes MTs to direct secretory vesicles and exocytosis to cell-matrix adhesion sites.^[18,19]

2.2. The tubulin GTP hydrolysis cycle drives microtubule dynamic instability

MTs grow by addition of GTP-tubulin onto existing MT plus ends. Because of the central importance and structural complexity of growing MT ends, much research has been directed toward understanding the biochemical and molecular details of MT polymerization and dynamic instability. Guanosine nucleotides are essential co-factors in both α- and β-tubulin, and two are present in each α/β-tubulin dimer. One GTP is trapped between the α- and β-tubulin moieties of the same dimer. Because this non-exchangeable or N-site has no GTPase activity, it is always GTP-loaded. In contrast, the other nucleotide binding site – referred to as E-site – is located at the β-tubulin end of the tubulin dimer, and in solution readily exchanges GDP for GTP. Free tubulin dimers, however, have essentially no GTPase activity because the catalytic E-site is incomplete.

E-site GTPase activation requires contacts with the next tubulin dimer in a protofilament because the adjacent α-tubulin provides an essential residue (E254) to the E-site catalytic pocket (Fig. 2a).^[20] In β-tubulin, this catalytic glutamate is absent explaining why the N-site does not hydrolyze GTP.^[21] GTP hydrolysis can thus only occur in the penultimate β-tubulin and consequently at least the last tubulin dimer in each protofilament at a growing MT end remains GTP-loaded unless it is lost by dissociation. It has long been postulated that this GTP-tubulin cap keeps MTs in the polymerizing state and protects MTs from transitioning to depolymerization (i.e. catastrophe).^[22,23] The free energy released by GTP hydrolysis is stored within the GDP-MT lattice,^[24] which primes MTs for rapid depolymerization once the protective cap is lost. Thus, although β-tubulin GTP hydrolysis is not directly required for MT polymerization, it does provide the energy driving MT dynamic instability. At typical intracellular GTP concentrations of around 0.5 mM, the vast majority of tubulin dimers in the cytoplasm are likely GTP-bound and thus polymerization competent.

Figure 2:

The MT plus end GTP-hydrolysis cycle. (a) The β-tubulin GTPase is not active in free tubulin dimers and the terminal subunit in a protofilament because completion of the E-site catalytic activity requires a specific glutamate residue (E254) in the next α-tubulin. GTP hydrolysis is thought to occur rapidly in the penultimate tubulin dimer, while random phosphate (P_i) release is slower resulting in a gradual but stochastic decrease of GDP·P_i tubulin dimers with increasing distance from the MT end. (b) Structural analysis suggests that completion of the β-tubulin E-site GTPase pocket also requires straight lineup of adjacent tubulin dimers in a protofilament. Thus, curled protofilaments at the end of a growing MT would be expected to contain mostly GTP-loaded tubulin dimers.

Once added to the MT end, the transition from GTP- to GDP-bound tubulin proceeds through at least two distinct biochemical steps (Fig. 2): The first is GTP hydrolysis itself after which both GDP and inorganic phosphate are still bound to β-tubulin (GDP·P_i), the second is release of the free inorganic phosphate. Bulk microtubule assembly biochemistry determined that it takes several seconds between GTP-tubulin incorporation and phosphate release.^[25] Thus, at physiological growth rates, the GDP·P_i zone trailing the growing MT end is expected to be several hundred tubulin dimers deep. In contrast, the GTP cap is estimated to be only one or two layers of tubulin dimers^[22,26] indicating that GTP hydrolysis is much faster than phosphate release. It follows that most tubulin dimers near growing MT ends must be in the GDP·P_i state. Consistent with the idea that GDP·P_i tubulin contributes to the protective cap, loss of the GDP·P_i zone at MT ends predicts catastrophe timing,^[27] and earlier work shows that GDP/BeF₃^-, a mimic of the GDP·P_i-bound tubulin state, protects MTs against catastrophes.^[28]

A relatively simple biochemical ‘coupled-random’ model in which the last tubulin dimer in each protofilament has no GTPase activity, but GTP hydrolysis probability is the same for all other tubulin dimers in the MT wall, predicts many aspects of MT dynamic instability.^[29] Because the GTP hydrolysis and phosphate release rates are independent of the tubulin dimer incorporation rate, faster growing MTs have longer protective caps. It thus becomes less likely for protofilaments to lose the cap. Consequently, there should be an inverse correlation between growth rate and catastrophe frequency, which is indeed observed with pure tubulin. However, other aspects of MT dynamic instability such as the apparent aging of growing MT ends – older ends are more likely to undergo a catastrophe – require a detailed understanding of the complex temporal evolution of the growing MT end structure.^[30]

2.3. GTP hydrolysis drives tubulin conformational changes

Previously reserved to X-ray crystallography, modern cryo-electron microscopy (cryo-EM) methods are approaching atomic resolution and beginning to reveal the MT lattice structural changes associated with biochemical transitions. The first such comparison of mammalian MTs in the GDP-state with MTs assembled in the presence of GMPCPP, a very slowly hydrolysable GTP analog that effectively mimics GTP-tubulin, found that the major difference between the two structures is a small change in length of the E-site interface between tubulin dimers. The interpretation of these data is that during GTP hydrolysis tubulin dimers along a protofilament move closer together by approximately 2.4%, but the shape of the tubulin dimer itself does not change much.^[31,32] This E-site compaction is also consistent with older measurements of the tubulin repeat length in super-twisted 14-protofilament MTs in which protofilaments wrap around the MT axis, which results in characteristic repetitive moiré patterns in electron microscopy images.^[33] More recent cryo-EM studies now also include GDP·P_i hydrolysis intermediates of MTs assembled either with GTPγS, another GTP analog that is thought to mimic the GDP·P_i state,^[34] or by means of the neuronal MT-binding protein doublecortin (DCX) as a nucleator and stabilizer of nascent polymerizing MTs,^[35] which allowed direct identification of MT structures in different nucleotide states. While both approaches have caveats – GTPγS is not GDP·P_i and we do not know how DCX-binding might alter MT conformation in these experiments – this is the best we have at this point and both studies indicate that E-site compaction occurs during the initial GTP hydrolysis step before phosphate release, accompanied by other subtler rearrangements of MT lattice geometry. Curiously, while E-site compaction has been observed systematically in mammalian MTs, it does not occur in the same way in yeast,^[36,37] indicating that GTP-hydrolysis associated changes of E-site conformation are not fully understood.

3. Straight or curly: The elusive structure of growing microtubule ends

Many biologically important interactions occur near growing MT ends. Yet, the precise structure of MT plus ends is difficult to decipher. It is not straight-forward to translate atomic detail into the mesoscale of a growing MT end where multiple molecular events and geometric constraints converge. Unlike the defined helical MT lattice, MT ends in cells and in vitro are heterogeneous.^[38–40] Thus, cryo-EM class averaging approaches that are so successful in boosting detail in geometrically repetitive structures do not work, and structural models must rely on observations of individual MT ends. However, because of the stochastic nature of MT dynamic instability, it is inherently difficult to know the polymerization state of a given MT end in a static image.

3.1. In vitro, growing microtubules can terminate in elongated sheets

Groundbreaking cryo-EM of in vitro MT polymerization reactions in conditions in which the majority of MTs is expected to be either polymerizing or depolymerizing initially found substantial differences between the end structures of growing and shortening MTs. Growing MTs appeared to have either blunt or tapered ends in which a subset of adjacent protofilaments is longer than the others.^[41–44] Fluorescence imaging and mathematical models of MT aging are consistent with short tapers at growing MT ends.^[45–47] However, some of the tapers observed in cryo-EM studies are very long with slightly curved and laterally connected protofilaments sometimes extending more than a micrometer. This inspired a model in which MTs grow as sheet-like extensions that subsequently close into tubes as the MT elongates.^[43] However, a mechanism how such tube closure might occur remains unresolved, and the observed curvature of such sheets is highly variable and does not seem to be specific to the GTP state.^[44] In contrast, cryo-EM images of rapidly depolymerizing MTs display a characteristic peeling of highly curled protofilaments away from the MT lattice.^[42,43] This apparent difference in protofilament curvature provided an attractive interpretation for how MT dynamic instability is powered that has been prevalent in the field for the last two decades: GTP-loaded tubulin dimers form relatively straight protofilaments that fit well into the MT wall, while intrinsically curved GDP-tubulin protofilaments introduce mechanical strain into the lattice that drives rapid depolymerization by outward protofilament peeling. Although uneven shortening of the interdimer interface could in principle contribute to curvature differences between GTP- and GDP-tubulin protofilaments, new ultrastructural analysis of different nucleotide state lattices indicates that reality is more complicated. In addition to the apparent strengthening of longitudinal contacts between tubulin dimers through E-site compaction, subtler differences between different guanosine nucleotide state MT lattices involving both shifts and twists of tubulin dimers along the protofilament axis may weaken lateral interactions between protofilaments during both GTP hydrolysis and phosphate release.^[35] Interestingly, the MT lattice seam appears as a weak point in the helical MT geometry, and although the physiological relevance of the MT seam remains unclear, it is reasonable to assume that it may act as a longitudinal fault line that comes apart first during the transition from MT growth to shortening.^[34]

3.2. In cells, most microtubule plus ends have curls

So, what does a growing MT end truly look like in physiological conditions? Even decades of research have not resolved this controversy, and several lines of evidence remain incompatible with MT growth as tapered elongated sheets. X-ray crystallography of GTP-tubulin dimers in different types of protein complexes all show high intrinsic curvature,^[48,49] and extended protofilament sheets are rarely seen in cells^[50]. A new and thorough quantitative analysis of MT end structures revisits this question by using state-of-the-art electron tomography of both chemically and cryofixed specimens to reconstruct MT ends in three dimensions.^[51] This work considers many controls and potential artifacts and focusses on different MT populations in cells and in vitro, concluding that there is no significant structural difference between the ends of growing and shortening MTs. Thousands of tracings show individual protofilaments bending outward, away from the MT wall in all conditions examined, and the high variability in curvature indicates protofilament flexibility perpendicular to the MT wall. Remarkably, in these cryo-EM tomograms, protofilaments are not curved in the other direction, indicating that they are more rigid parallel to the MT wall. Flexibility in one direction and rigidity in the other is also consistent with the narrow curvature distribution of in vitro assembled protofilament rings.^[52] At a growing MT end, this implies that outward curved protofilaments must splay away from the MT axis and cannot interact laterally without straightening first (Fig. 3). Because efficient GTPase activation likely requires protofilament straightening beyond Brownian fluctuations,^[35] we expect these curled protofilaments to be mostly GTP-tubulin at a growing MT end (Fig. 2b). The average length of about five tubulin dimers in each of these curled protofilaments matches the biochemical estimates of less than fifty MT end GTP-tubulin subunits remarkably closely.^[22]

Figure 3:

Model of plus end structure and guanosine nucleotide distribution in a slow- and a fast-growing MT. MT growth requires two coordinated processes: elongation of individual protofilaments by addition of GTP-tubulin subunits and lateral closure of protofilaments into a tube as indicated in the end-on view. Also note the spatial separation of protofilament tips due to their lateral rigidity, and thus the absence of ‘corner’ sites. The geometry of the curled protofilaments is based on recent quantitative electron tomography data.^[51] At a given tubulin dimer addition rate, we predict that faster protofilament zippering would result in shorter protofilament extensions in a faster growing MT. The guanosine nucleotide distribution at the indicated MT growth rates was calculated by stochastic simulation using published rate constant estimates.^[67] Because these rate constants are based on the maturation and turnover of EB1 binding sites, they only indirectly reflect GTP hydrolysis. We assume here that GTP is only hydrolyzed once tubulin is integrated in the MT wall and not in the curled protofilament extensions, and we may therefore overestimate the length of the GTP zone. Note the extensive mixed lattice of GDP and GDP·P_i tubulin dimers especially in the fast-growing MT.

Overall, these new tomography findings are consistent with earlier studies in which most growing MTs in cells had flared or funnel-shaped ends that terminate in outward curved protofilaments^[38,39,53], but data from metazoan and mammalian cells are still very sparse. Nevertheless, with these new data and improved technology it seems difficult to deny that MT ends in cells are structurally similar notwithstanding their polymerization state. The reason for the structural differences observed by different research groups especially in seemingly equivalent in vitro MT polymerization reactions remains unresolved and is puzzling. It is possible that contaminations or inactive tubulin dimers in vitro slow elongation of subsets of protofilaments resulting in uneven MT growth and sheet-like extensions. Alternatively, it might well be that the MT end structure is intrinsically highly variable, and it is reasonable to assume that mechanisms have evolved to coordinate the different biochemical and structural processes associated with MT assembly to optimize MT growth, resulting in a more homogenous appearance of MT ends in cells.

Our knowledge of how ultrastructural transitions are linked to the biochemical state of tubulin dimers near growing MT ends is based on the high-resolution structural data of homogenous assembled MT lattices in different guanosine nucleotide states. This limits our understanding as these do not truly represent the non-equilibrium dynamics of growing MT ends. If there are no strong cooperative effects between adjacent protofilaments, stochastic GTP hydrolysis with realistic biochemical rate constants and MT growth rates produces an extended region in which the MT lattice near the growing end is expected to be highly heterogenous and different tubulin guanosine nucleotide states are mixed (Fig. 3). How tubulin dimers having potentially incompatible conformations can exist side-by-side, and therefore what conformation individual tubulin dimers adopt in such mixed lattices remains unclear. Findings that GTP- and GDP-loaded tubulin can co-polymerize^[54] and that GTP-tubulin can repair GDP-MT lattices^[55] indicate that there must be a great deal of conformational flexibility, but which conformation is the most stable in a mixed lattice is not at all clear. GDP-tubulin incorporation also indicates that the protective biochemical cap does not operate on the single protofilament level but must be a collective property of the growing MT end structure.

4. Twists and supertwists: Molecular recognition of protofilament conformation

4.1. End-binding proteins recognize the transient GDP·P_i tubulin conformation

Another puzzle piece in understanding MT plus end structure is contributed by proteins that bind to growing MT ends. These +TIP proteins were discovered in the late 1990s because of their unusual intracellular dynamics.^[56,57] End-binding (EB) proteins – with EB1 being the most abundantly expressed in mammalian cells – are the central components of a +TIP protein network.^[58] Only EBs directly recognize a tubulin conformation specific to growing MT ends. Treadmilling of this biochemically distinct zone as MTs grow results in the characteristic appearance of +TIP comets moving through the cytoplasm. Although high affinity EB-binding sites exist for several seconds, the dwell time of individual +TIP molecules is 10–100-fold shorter.^[59–62] Thus and importantly, the apparent +TIP comet movement reflects the turnover of the underlying EB binding sites at growing MT ends, and no actual transport of +TIP protein molecules occurs.

Cryo-EM ultrastructural analysis of EB-decorated MT lattices demonstrated that EBs recognize a site located longitudinally and laterally between protofilaments at the interface of four tubulin dimers (Fig. 4).^[63] Again, using MTs assembled with different guanosine nucleotide analogues these experiments also showed that EBs bind only weakly to GDP MTs, but have a high affinity for MTs assembled with GTPγS or GDP/BeF₃^-, both thought to represent the GDP·P_i transition state.^[64,65] A recent clever approach examining EB-binding to different guanosine nucleotide state zones on the same MT confirms this preference for the GDP·P_i lattice for all mammalian EB paralogues, and also demonstrates that all EBs have very low affinity for GMPCPP (i.e. GTP-like) MT lattices.^[66] In these experiments, transitions of EB-binding between these different zones are very sharp. This is unlike the shallow exponentially decaying profile of EB comets on fast growing MT ends in cells, which can be explained by the first order reaction kinetics of phosphate dissociation,^[58,67] and provides further support for a mixed guanosine nucleotide state trailing growing MT ends. EB comet splitting in the presence of certain MT-targeting drugs also supports that different guanosine nucleotide MT lattice states can exist in adjacent protofilaments.^[68]

Figure 4:

Molecular recognition of MT lattice conformations. (a) EBs and DCX recognize the same location in between four tubulin dimers in different guanosine nucleotide states. DCX binds to the GDP MT lattice and has been proposed to bind cooperatively.^[74] EBs recognize the GDP·P_i MT lattice. DCX has two MT-binding domains and EBs are obligatory dimers. Yet, it is not known for either molecule how the connected MT-binding domains are arranged on the MT lattice. The only available structural data for EB dimers spaces the MT-binding domains at ~5 nm,^[97] which can span the distance laterally across protofilaments. However, flexibility of the interdomain linker in EBs does not preclude binding longitudinally between adjacent protofilaments. Both configurations would stabilize lateral protofilament interactions. Note the absence of EB- (or DCX-) binding sites along the MT lattice seam. (b) Human breast adenocarcinoma cell expressing fluorescently tagged DCX and EB1. Note the mutually exclusive zones of EB- and DCX-binding at growing MT ends (inset).

Each EB binding site is flanked by two GDP·P_i β-tubulins. However, in a mixed lattice at a growing MT end, only around a third of potential EB-binding sites will have two GDP·P_i tubulin dimers next to each other. While this clearly represents the best EB binding site, different affinities of different EBs for mixed GTP / GDP·P_i or GDP·P_i / GDP binding sites could explain the different intensity profiles of the three mammalian EBs observed in cells.^[66] However, EBs are obligatory homo- or heterodimers adding substantial complexity to possible binding site combinations.^[69,70] Because the EB binding site requires lateral contacts between protofilaments, it does not exist on curled protofilaments that splay apart at the MT end. This is consistent with observations that EB1 comets do not extend to the very end of growing MTs.^[67,71] Preferential EB binding to the GDP·P_i transition zone rather than the terminal GTP tubulin subunits also resolves discrepancies between biochemical estimates of relatively few GTP tubulin dimers at MT ends and the extended hundreds of tubulin dimer long EB binding zone,^[72] and because of the independence of phosphate release and tubulin addition rates faster growing MTs have longer GDP·P_i zones and thus longer EB comets.

4.2. Doublecortin recognizes the GDP microtubule lattice

Are there other proteins that recognize specific MT guanosine nucleotide states? Remarkably, DCX, a neuronal protein that is structurally unrelated to EBs, binds to the same site between tubulin dimers.^[73] However, even though at very low concentrations DCX associates with growing MT ends in vitro,^[74] quantitative microscopy in cells demonstrated that DCX binds along the MT lattice and is non-competitively excluded from the EB zone,^[75] which indicates that EBs and DCX recognize different MT-lattice nucleotide conformations. In contrast to EBs, DCX is thus the only known example of an ‘anti-+TIP’ with high affinity for the GDP MT lattice (Fig. 4). Consistent with DCX exclusion from the zone near the growing MT end, DCX does not influence MT growth rates, but does decrease the rate of MT shortening.^[75]

Based on the striking dependence of DCX-binding on MT geometry – DCX only binds straight, but not curved MTs in cells^[75] – and the reversal of this specificity in the presence of taxanes,^[65,75] MT-stabilizing drugs that appeared to reverse E-site compaction,^[31] we initially proposed that DCX can sense the state of E-site compaction as tubulin dimer repeat length would change in curved MT segments. However, newer data show that E-site compaction has already occurred in the GDP·P_i MT lattice^[34,35] and the effect of taxanes on E-site compaction has become less clear.^[35,76] Instead, different guanosine nucleotide MT lattices display different protofilament twists in which tubulin dimers along a protofilament are slightly rotated relative to each other. These protofilament twists resolve into MT lattice supertwists in which protofilaments wrap around the MT axis at a shallow angle to preserve lateral tubulin dimer interactions. Bending of a supertwisted MT lattice should affect protofilament twists on the inside and the outside of the curved segment in opposite directions. This may distort the geometry of the DCX binding site and provide an alternative explanation for DCX sensitivity for MT curvature. The strong preference of DCX for 13-protofilament MTs over other geometries with different supertwists is in principle consistent with this idea.^[74,77]

Similarly, EBs cannot simply recognize E-site shape as both GDP and GDP·P_i MTs have compacted lattices. Instead, EBs may recognize the small left-handed twist found in GDP·P_i-mimicking GTPγS MT lattices that runs in the opposite direction to the dimer twist in GDP- or GTP-like MTs.^[34] Indeed, EBs enhance the left-handed protofilament twist in GTPγS MTs, which may allosterically increase EB- and decrease DCX-affinity, and also stimulate GTP hydrolysis or phosphate release.^[32,34,64] However, it should be noted that the twists in different guanosine nucleotide MT lattices are very small,^[78] and a full turn of a skewed protofilament around a MT requires tens of micrometers.

4.3. CKAP5 localizes to protofilament tips

CKAP5 is a processive tubulin polymerase of the TOG/XMAP215 family that localizes to growing MT plus ends and increases the MT growth rate in vitro.^[79,80]. CKAP5 likely recognizes a structural feature only accessible at the very end of protofilaments and is not known to bind specific tubulin guanosine nucleotide states. X-ray crystallography, however, shows TOG-domain association with curved tubulin dimers.^[81,82] Thus, CKAP5 may recognize curled protofilaments on growing MT ends, which fits well with the distinctive punctate CKAP5 localization to the extreme tip of growing MTs in front of the EB domain both in cells and in vitro (Fig. 5).^[67,71] Experiments employing conformation-changing tubulin mutations further suggest that curled protofilaments are indeed required for CKAP5 to processively hold onto growing MT ends.^[83] CKAP5 TOG domains are thought to bind and tether tubulin dimers to growing MT ends increasing the local concentration and thus the rate of tubulin dimer incorporation,^[82,83] which may involve unrolling of the compact TOG-domain array as tubulin dimers are added to growing protofilament ends.^[84] Earlier measurements showing CKAP5-induced elongation of MTs by steps greater than the length of individual tubulin dimers suggest that multiple tubulin subunits may be delivered together.^[85]

Figure 5:

Model how cells achieve optimal MT growth rates assuming acceleration of two independent structural transitions: Protofilament elongation catalyzed by CKAP5, a MT polymerase of the TOG/XMAP215 family, and subsequent closure of lateral protofilament interactions by EBs and associated +TIPs. CKAP5 alone increases the growth rate (v_g) by accelerating tubulin dimer addition to protofilament ends. However, this becomes unproductive if downstream structural transitions that lead to tube closure cannot catch up. In contrast, EBs alone have little effect on protofilament elongation, but enhanced protofilament straightening and tube closure results in a loss of the protective GTP/GDP·P_i cap and more frequent transitions from growth to shortening (f_cat; catastrophe frequency). DCX binds to the GDP MT lattice and is excluded from the growing MT end. Consequently, DCX has no effect on growth rate, but inhibits MT depolymerization (v_s; shortening rate). These predicted effects on MT dynamics are supported by experimental results. Note that the guanosine nucleotide state and protein-binding zones are not drawn to scale here (see Fig. 3) and that conformational changes of tubulin dimers (E-site compaction and protofilament twists) are not included because of their incompletely understood relevance for MT growth dynamics and their small scale relative to the size of assembled MTs.

5. Tubulin addition and protofilament zippering: Optimized for rapid MT growth

In addition to frequent transitions between growing and shortening phases, a striking characteristic of MTs in many metazoan cells is a much higher and more variable growth rate compared with in vitro MT polymerization or even MTs in yeast. The average growth rate of an interphase MT in the interior of mammalian cells is around 20 μm/min with bursts of up to twice that rate,^[86,87] while purified tubulin in vitro even at high concentrations does not exceed growth rates of 1–2 μm/min and shows little variation. At estimated 10 μM intracellular tubulin concentrations, tubulin subunit addition to growing MT ends is not diffusion limited and diffusion alone should easily support growth rates >50 μm/min.^[88] Thus, MT growth is evidently more complex than simple tubulin incorporation.

Combining biochemical and structural data across scales, a modern model of a growing MT end must include at least three distinct but overlapping zones (Fig. 3): 1) Short, curled protofilament extensions at the distal end that splay outward away from the MT axis; 2) an exponentially decaying zone of GDP·P_i tubulin subunits that constitutes the binding platform for EBs and +TIPs; and 3) the remainder of the MT mostly composed of GDP tubulin although islands of GTP-like conformation may result from severing and repair processes.^[55,89,90] In such a model, addition of tubulin dimers to protofilament ends and zippering of the lateral interactions between protofilaments could both be rate-limiting to how fast a MT can grow (Fig. 5). If protofilament zippering does not catch up with GTP-tubulin addition, stimulation of protofilament elongation – for example through a MT polymerase such as CKAP5 – becomes non-productive and would not lead to faster elongation of the MT tube. Conversely, if GTP-tubulin dimers are not incorporated into protofilament ends fast enough, closing up protofilaments into a tube – for example through EB proteins promoting lateral interactions – cannot increase the tubulin addition rate. Consequently, in vitro MT growth stimulation by CKAP5 alone saturates at growth rates well below what is observed in many eukaryotic cells,^[79] and EBs by themselves barely affect the MT growth rate even at high concentrations.^[91] In vitro, EBs however promote catastrophes, which is consistent with loss of the protective cap when tube closure and associated GTP hydrolysis is faster than protofilament elongation.^[67,91] EBs also directly increase the tubulin GTP hydrolysis rate.^[32] This model of course ignores the actions of many other cellular factors that can control aspects of MT dynamics, but it is striking that in the presence of CKAP5 and EBs together, in vitro growth rates at physiological tubulin concentrations closely approach rates observed in cells beyond the effects of the individual proteins.^[92] This synergy between CKAP5 and EBs has been attributed to possible long-range allosteric effects of EBs on the MT lattice. However, such allostery is not strictly required, and the observed synergy could alternatively be explained by two rate-limiting processes that are independently accelerated by either CKAP5 or EBs.

While this explains in vitro steady-state MT polymerization quite well, the situation in cells is more complicated and less well understood. As expected, both removal of CKAP5 or EBs decreases intracellular MT growth rates.^[87,93,94] However, in cells CKAP5 localization to MT ends does not correlate well with the MT growth state: MT end-associated CKAP5 dots are often strongest on MTs that are starting to grow and display very weak EB comets, while fast-growing MT ends with pronounced EB comets frequently have no detectable CKAP5.^[87] A possible explanation for this observation is that rapidly growing MT ends in cells oscillate between fast-growing blunt ends with long GDP·P_i zones and slower growing states with more pronounced curled GTP protofilament caps.

In contrast to in vitro, simple stabilization of lateral protofilament interactions by the EB MT-binding domain itself is insufficient to rescue MT growth defects in cells, but requires other +TIPs recruited to the MT end zone by EBs.^[87] It is unclear why this is the case, but in cells, additional structural transitions at growing MT ends may become relevant. For example, in the crowded and viscous cytoplasm thermal fluctuations alone may not be sufficient and protofilaments may have to be actively straightened before they can laterally interact. CKAP5 and EB activities also must communicate to maintain a productive equilibrium of the biochemical and structural changes at growing MT ends to achieve rapid persistent growth, an important functional property of the intracellular MT network. EB-recruited +TIPs such as SLAIN2 are candidates for these activities, but the underlying mechanisms remain ill-defined.^[94,95] It should be noted that even in the absence of CKAP5 or EBs, the MT growth rate in cells is still substantially higher than in vitro. While this may partially have to do with limitations in accurately tracking slow-growing MTs in the cell interior or molecular crowding effects of cytoplasm,^[96] evidently other factors contribute to controlling MT growth in cells.

6. Conclusion

Supported by new quantitative data indicating that the tip of growing MTs is composed of individual outward curled protofilaments rather than laterally connected sheets, we propose what we believe is a realistic view of growing MT ends. Even though many details remain uncertain, such as for example the nucleotide state of these curled protofilaments, the MT end is evidently a highly dynamic structure with distinct but partially overlapping biochemical and structural zones. These zones are recognized by distinct proteins and allow for multiple levels of regulation. It should also be noted that it is likely that there is considerable heterogeneity in growing MT end structures in cells. In addition, detailed analysis so far has focused on in vitro, yeast or mammalian MT populations that grow comparatively slowly, and we have yet to see the end structures of rapidly growing MTs in the interior of mammalian cells.

The presence of curled protofilaments at growing MT ends also has important topological consequences. For example, corner sites that were thought to enhance tubulin incorporation by simultaneously allowing longitudinal and lateral interactions are basically non-existent in flared MT ends in which individual protofilaments ends are likely separated by 10 nm or more. This suggests that there is no coordination between elongation of neighboring protofilaments. It will be exciting to see how these new structural details will be incorporated into more-and-more realistic mechano-chemical models of MT dynamics.^[46]

Finally, we postulate that the rapid physiological MT growth observed in metazoan cells could be achieved by optimized control of independent and potentially rate-limiting transitions at growing MT ends: protofilament elongation and lateral protofilament zippering that results in closure of the MT wall. We predict that optimal, fast MT growth is achieved when these transitions are balanced, and MT tube closure occurs at rates similar to protofilament elongation preventing non-productive buildup of long protofilament extensions. In cytoplasm, this may require a physical link between the proteins catalyzing these transitions. To test this hypothesis, it will be necessary to quantitatively compare the structures of growing MT ends in the presence of CKAP5 and EBs and correlate MT end structures with growth rates. It seems that the experimental tools are available to address this question. Evidently, recent years have seen enormous advances in our understanding of MT structure and dynamics. However, much remains to be learned and we think that – pun intended – the end is not yet in sight.

Abbreviations:

cryo-EM

cryo-electron microscopy

EB

end-binding

MT

microtubule

+TIP

MT plus end-tracking protein

DCX

doublecortin

CKAP5

cytoskeleton-associated protein 5

GTP

guanosine triphosphate

GDP

guanosine diphosphate

GMPCPP

Guanosine-5’-[(α,β)-methyleno]triphosphate

GTPγS

Guanosine-5’-(γ-thio)-triphosphate

P_i

inorganic phosphate