Dynamic Microtubules as Sensors in Animal Cells

Abstract

Microtubules physically organize eukaryotic cells by serving as structural elements and polarized transport tracks. This article advances the hypothesis that dynamic microtubules also serve as sensors of cell shape and cytoplasmic state, building on ideas proposed for higher plant cells¹. Microtubule polymerization dynamics and lattice structure are sensitive to mechanical, chemical and signaling inputs which alter the balance between microtubules and soluble tubulin and regulate MAP binding affinity. These changes are detected by transducers which include the GTP exchange factor GEF-H1 (ARHGEF2) and MARK family kinases. The resulting signals regulate cytoplasmic behavior, gene expression and tissue physiology. The microtubule destabilizing drugs colchicine and plinabulin may mimic sensing of pathophysiological cues by microtubules, leading to activation gene expression programs that promote cell survival, growth and repair which account for the therapeutic actions of the drugs. In tissue cells with stable morphologies, the sensory functions of microtubules may be as or more important than their architectural functions. This re-framing of microtubule biology suggests new directions for mechanistic inquiry and drug discovery.

Introduction

Microtubules are long, stiff protein rods that physically organize eukaryotic cells by serving as structural elements and polarized tracks for organelle transport. In many cell types, they undergo rapid polymerization dynamics on free plus ends through a process called dynamic instability, whereby individual ends alternate between bouts of growth and shrinkage fueled by hydrolysis of GTP bound to β-tubulin². These dynamics give rise to length fluctuation of individual microtubules on scales of microns and seconds. Rapid polymerization dynamics are necessary for reorganization of microtubules when cells change shape or enter mitosis, but their purpose in tissue cells with stable interphase morphologies is unclear. This article explores the hypothesis that dynamic microtubules play a sensory role. There are precedents for this hypothesis; it was proposed for abiotic stress sensing in plants in a prescient review¹ and the actin cytoskeleton is known to sense substratum stiffness as well as generate shape and force³. Consideration of sensory functions can rationalize aspects of microtubule biology that are puzzling from a purely cell architecture viewpoint, including the biological purpose of signaling proteins that are modulated by microtubule assembly state and the therapeutic activity of microtubule destabilizing drugs such as colchicine and plinabulin.

Borrowing from Webster’s dictionary, a sensor is here defined as “a device that responds to a physical or chemical stimulus (input) and transmits a resulting impulse for operating a control (output)”. This article will consider inputs that change the stability and lattice structure of microtubules and outputs that modulate signaling and gene expression (Figure 1). This generalized model prompts the need to identify physiological and pathophysiological inputs, signal transduction modules and outputs that modulate tissue biology.

Figure 1. General model for sensing by dynamic microtubules.

Inputs change the balance between microtubules and soluble tubulin or the conformation of the microtubule lattice. Inputs that change post-translational modifications could also be included. Signal transducers respond to microtubule assembly state or lattice conformation and convert that information into signaling pathway modulation. Outputs are changes in cell and tissue physiology mediated by signaling and gene expression.

Related topics discussed elsewhere

Alternative hypotheses for the purpose of microtubule dynamics in morphologically stable cells include a default state of the system, a requirement to maintain steady states in geometric organization and promoting vesicle capture for dynein-mediated transport⁴. The response of mitotic cells to microtubule perturbation is mediated by the Spindle Assembly Checkpoint (SAC)⁵. Stable microtubule bundles play important roles in sensory biology, notably as the physical organizers of non-motile sensory cilia in vertebrate photoreceptors and invertebrate chemosensory neurons ⁶,⁷. Primary cilia extend from the surface of manyx cell types where they serve as chemo- and mechano-sensors⁸. Important signaling pathways such as hedgehog depend on kinesin and dynein-mediated transport of receptors along primary cilium microtubules⁹. Bundles of stable microtubules are also important for touch sensing by specialized neurons¹⁰. In these examples, physical and chemical stimuli are sensed by plasma membrane spanning ion channels or chemoreceptors while stable microtubules serve as architectural elements. In this article, the focus is on pathways where dynamic microtubules themselves are the sensory element.

Origins of the sensor hypothesis

In addition to Nick’s review¹, this article was stimulated by a failed prediction. After discovering dynamic instability, Marc Kirschner and I predicted that microtubules would be highly dynamic during morphogenesis but would stabilize when cells achieved a stable tissue morphology¹¹. We now know from EB1/3 comet imaging that most tissue cells retain dynamic microtubules⁴,¹²,¹³,¹⁴ with some exceptions¹⁵. Polymerization dynamics-promoting proteins such as EB1 (MAPRE1) and CH-TOG (CKAP5) are broadly expressed in single cell mRNA atlases, suggesting that most cell types contain dynamic microtubules. This prompts the question, what are growing plus ends doing as they course through the cytoplasm of morphologically stable tissue cells? Could they be making measurements? This article was also stimulated by my interest in the therapeutic mechanisms of microtubule-targeting drugs. The anti-inflammatory drug colchicine depolymerizes microtubules in hepatocytes and induces a gene expression program which is responsible for its therapeutic activity¹⁶. This prompted the questions, what is the biological purpose of the colchicine response, and what liver biology does colchicine treatment mimic? The sensor hypothesis provides a conceptual basis for addressing these questions.

Physiological and pathophysiological inputs

Mechanical:

The length and stiffness of microtubules, and their tendency to grow from the center of the cell towards the periphery, make them well-suited to sensing cell shape. Growing plus ends usually switch to shrinking when they touch the cortex at the periphery of the cell, which is the first step in mechanosensing. If the microtubule organizing center (MTOC) is positioned away from the cell center, the resulting asymmetries in microtubule length and density generate forces that move it towards the cell center (Figure 2A). MTOC centering can occur by a variety of physical mechanisms¹⁷ and plays a central role in the organization of simple cells, zygotes and early embryos¹⁸. MTOC centering is a good example of mechanosensing by dynamic microtubules, but the focus of this article is on more complex mechanisms which regulate signaling and gene expression. Based on current literature, one important way that dynamic microtubules regulate signaling in response to cell shape is pathway activation by local absence of microtubules (Figure 2B). This non-obvious mechanism is revealed by transducer mechanisms discussed below. Physical damage, cell growth and morphogenesis can all deform cell shape to generate microtubule-depleted regions at the cell periphery which activate signaling by microtubule absence. Figure 2C illustrates an alternative mechanism in which mechanical forces bend microtubules or induce conformation changes in the lattice which regulate downstream transducer proteins. Doublecortin (DCX) binds preferentially to curved lattices¹⁹ and Tau (MAPK) to compacted lattices²⁰, suggesting that signal transducers may respond to lattice modulation. Microtubule turnover by dynamic instability refreshes the lattice and is expected to terminate signaling downstream of lattice changes, which would make signal duration sensitive to turnover rate in this model.

Figure 2. Mechanosensing of cell shape by dynamic microtubules.

A. MTOC centering by physical mechanisms. Asymmetries in the length and density of microtubules (green lines) generate forces that center the microtubule organizing center (brown dot). B. Signal activation by microtubule absence. Deformation of the cell generates microtubule-depleted regions (red hatching). These activate pathways which signal back to the nucleus (red arrow). C. Signal activation by alteration of microtubule lattice structure in response to physical forces. The cartoon shows lattice bending by compression. Lattice expansion by tension is another way the lattice could change.

Chemical:

Microtubule dynamics can be poisoned by structurally diverse ligands which bind to at least five sites on beta tubulin²¹ as well as by cysteine-modifying electrophiles²². Electrophiles with this potential are common in plant and fungal secondary metabolites²³ and are also generated endogenously by oxidative stress²⁴. The chemosensitivity of tubulin could reflect an accident of its fold or, more speculatively, an evolved chemosensory function. In either case, polymerization dynamics can have a chemosensory function. The gut is the first line of defense against toxic phytochemicals. Diarrhea is a common side effect of the microtubule destabilizing drug colchicine which is usually attributed to its anti-mitotic activity²⁵. An equally plausible hypothesis is that gut chemosensory neurons sense colchicine via microtubule depolymerization and trigger diarrhea as a response, with the purpose of expelling toxins from the gut. Once colchicine enters the bloodstream, it depolymerizes microtubules selectively in hepatocytes¹⁶ (Figure 3). This triggers a cytoprotective gene expression program mediated in part by the transcription factor NRF2, the master regulator of defense against electrophiles and oxidants²⁶. NRF2 is negatively regulated by the well-validated chemosensory protein KEAP1, which is degraded when its reactive cysteines are alkylated or oxidized. Colchicine caused KEAP1 degradation in hepatocytes¹⁶, suggesting that tubulin serves as a parallel chemosensor.

Figure 3. The response of hepatocyte microtubules to colchicine suggests a chemosensory function.

The left panel shows a section of mouse liver stained for microtubules after vehicle treatment, the right 6hrs after administering a safe anti-inflammatory dose of colchicine (0.4mg/kg). See¹⁶ for details. The safe dose depolymerized microtubules in hepatocytes but left them undamaged in endothelial and Kupfer cells as well as circulating leukocytes (not shown). Depolymerization activated a hepatocyte gene expression program that was cytoprotective in the liver and anti-inflammatory in the whole animal via factors secreted into blood. By extension, hepatocyte microtubules may serve as sensors for other food-derived toxins and electrophiles which damage tubulin

Temperature:

Microtubule polymerization is famously cold sensitive. Nick and colleagues proposed that microtubules serve as thermometers for cold stress in plants, responding either via their intrinsic cold sensitivity or indirectly via upstream signals such as cytoplasmic calcium¹,²⁷. This hypothesis is part of a larger concept that both the actin and microtubule cytoskeletons sense stressful physical stimuli in plants²⁸. To what extent this hypothesis extends to animal cells is an interesting open question. The current consensus is that temperature is sensed primarily by ion channels in animals²⁹, but microtubules could play secondary roles. Tubulin isotype composition responds adaptively to cold in Drosophila³⁰, which could be a signature of temperature set point tuning. Deeper analysis is warranted into how microtubules adapt to, and perhaps sense, changes in temperature.

Infection:

Many viruses require microtubules for transport within host cells and some stabilize microtubules as part of their life cycle³¹. An emerging concept in innate immunity is that cells sense pathogen-induced perturbation of normal homeostatic processes and respond by activating defensive programs³². Perturbation of microtubule dynamics might be one such trigger. The microtubule-regulated transducer GEF-H1 (discussed below) is required for efficient innate immune responses to viruses and bacteria, working in concert with receptors for pathogen-specific molecules such as RigI and PGLYRP-1³³,³⁴. GEF-H1 may play a role in anti-viral responses downstream of microtubule perturbation. A pathogen sensing/response perspective might help explain the complex phospho-regulation of the axonal MAP Tau that has been mostly studied in the context of neurodegenerative disease but also changes during viral infection³⁵. Sensory neurons are sensitive to microtubule stabilization, as evidenced by neuropathic side effects when taxanes are used for cancer treatment³⁶. The usual explanation is that neurons are highly dependent on the transport function of microtubules³⁶. An alternative hypothesis is that neurons interpret microtubule stabilization as a sign of virus infection and mount an inflammatory response. Vinca alkaloids also cause peripheral neuropathy, whether by the same pathways as taxanes is unclear³⁷.

Signaling:

Dynamic microtubules can sense the activation state of signaling pathways and regulate downstream signal propagation. An interesting example is regulation of the barrier function of endothelial cells, which separate blood from tissue and gate the passage of water, proteins and cells. The inflammatory cytokine TNFa destabilizes endothelial microtubules, which in turn increases permeability of the endothelial barrier via GEF-H1 activation³⁸,³⁹. This suggests a role of microtubules in integrating barrier-regulating signals. In another example, paclitaxel inhibits S-phase entry triggered by growth factors in quiescent fibroblasts⁴⁰ which suggests a role for microtubules in integrating growth signals. In principle, the polymeric, highly charged structure of the microtubule lattice makes it well suited to participate in signal initiation, amplification and crosstalk. Multiple signaling proteins can dock onto microtubules, which can promote proximity-based enhancement of phosphorylation and other signaling biochemistry. In these respects, the microtubule lattice conceptually resembles patches of phosphorylated inositol lipids on the inner leaflet of the plasma membrane which generate and integrate signals from cell surface receptors. The best characterized example of signal amplification on microtubules is proximity-mediated auto-phosphorylation of Aurora-B kinase (part of the Chromosome Passenger Complex or CPC) when it binds at high density to midzone microtubules during cytokinesis⁴¹,⁴². Conceptually related biochemistry could regulate other kinases and GTPases.

Pathophysiology:

This is a catch-all for changes in tissue and cell physiology associated with disease states which perturb microtubule dynamics. One of the best characterized is axotomy (cutting an axon), which models physical trauma to the nervous system and causes initial depletion of microtubules followed by regrowth in peripheral neurons⁴³,⁴⁴. Microtubule damage in neurons has usually been viewed as a downstream consequence of disease states, notably in neurodegeneration⁴⁵. However, microtubules also play an active role in sensing pathological inputs and inducing protective responses⁴⁶ (discussed in more detail below).

Transducers of microtubule assembly state and lattice structure

The best characterized transducer is the GTP exchange factor GEF-H1 (ARHGEF2), a previously obscure member of the ARHGEF family which was shown to bind to microtubules resulting in suppression of its GEF activity⁴⁷ (Figure 4A). Subsequent work confirmed that monitoring the assembly state of microtubules is a central function of GEF-H1⁴⁸. Precisely how dissociation from microtubules activates GEF-H1 is unclear, and its activity also responds to other inputs⁴⁹. Once released from microtubules and activated, GEF-H1 activates and promotes membrane binding of the small GTPases RhoA, RhoB and perhaps RhoC. RhoA in turn activates ROCK leading to activation of Myosin-II, which causes cell rounding when microtubules are depolymerized⁵⁰. RhoB is less studied but plays an important and distinct role in many cell types, for example to regulate barrier function in endothelium⁵¹. Gene expression responses to microtubule depolymerization were greatly reduced in dendritic cells from GEF-H1^−/− mice compared to wt.⁵², though some regulation remained suggesting involvement of additional transducers. We found that inhibition of all Rho GTPases using a bacterial toxin completely blocked gene expression responses 6hrs after microtubule depolymerization in endothelial cells, while inhibition of ROCK had much less effect⁵³. These studies show that the main pathway from microtubule destabilization to rapid changes in gene expression go through the GEF-H1-Rho pathway but do not depend on ROCK.

Figure 4. Signal transduction downstream of microtubule depolymerization.

A.GEF-H1 pathway. The Rho GTP exchange factor GEF-H1 (ARHGEF2) is sequestered in an inactivate state when bound to microtubules. Destabilization causes it to dissociate and activate, leading to activation of Rho family GTPases and then to large changes in gene expression and myosin-II activation. GEF-H1 can also be activated by microtubule acetylation⁵⁵,⁵⁶.

B. MARK kinase pathway. MARK1–4 are evolutionarily conserved kinases which bind to microtubules and phosphorylate MAPs as well as regulating many other processes involved in cell polarization. Many MARK phosphorylation sites were lost after depolymerizing microtubules in endothelial cells⁵³. The biochemical mechanism is not known, as implied by the question mark. Decreased MAP phosphorylation in response to microtubule destabilization provides homeostatic feedback control of microtubule stability. Other signaling outputs are likely given the complex role of MARKs in cell polarity.

GEF-H1 is expressed in most cell types and new details of its regulation and functions are still emerging. It is instructive to read how different investigators conceptualized its biological purpose. Basic cell biologists implicated it as a major communication link between the microtubule and actin cytoskeletons⁴⁷ which coordinates these systems during spreading of the leading edge⁴⁹ and mediates microtubule-stimulated dissolution of focal adhesions at the rear of migrating cells⁵⁴. Vascular biologists implicated it in regulation of endothelial barrier permeability in response to inputs which perturb microtubules³⁹. Immunologists implicated it as a modulator of immune responses to infection and cancer⁵². The role of GEF-H1 in neurons and glia has not been systematically investigated, but it expressed in those cells and likely to play important roles in homeostasis and pathology. Recent studies showed that GEF-H1 transduces tubulin post-translational modifications into RhoA signaling. Increased tubulin acetylation on a lumenal site caused release and activation of GEF-H1⁵⁵,⁵⁶. Lumenal acetylation makes microtubules more flexible and is implicated in their response to bending⁵⁷, suggesting GEF-H1 transduces signals from both microtubule depletion (figure 2B) and deformation (figure 2C).

The Microtubule Affinity Regulating Kinases (MARK1–4 in humans) have long been known to regulate the activity of Tau-family MAPs which stabilize microtubules⁵⁸,⁵⁹. We recently implicated them as transducers of microtubule assembly state (Figure 4B). This new proposed function for MARKs was inferred from proteomics and has not yet been confirmed by mechanistic studies, so should be considered hypothetical. The MARKs were discovered biochemically as kinases that phosphorylate Tau (MAPT), MAP2 and MAP4 on conserved KXGS motifs in their microtubule binding repeats⁶⁰. This phosphorylation lowers the affinity of the MAPs for binding to microtubules in biochemical assays and destabilizes microtubules in cells. The C. elegans MARK homolog was discovered independently and earlier as PAR-1, a gene required to polarize the early embryo⁶¹. MARKs were subsequently implicated in cell polarization in multiple species⁶² where they phosphorylate many substrates and their kinase activity is regulated by many inputs⁶³. Using phosphoproteomics, we observed rapid loss of more than 150 phosphosites when microtubules were depolymerized by drugs in endothelial cells, of which approximately half were imputed to be MARK sites by motif analysis⁵³. Many, but not all, of these microtubule-dependent MARK sites are in proteins known to bind microtubules, including KXGS sites in the microtubule binding repeats of MAP4. Based on these data, I propose that the combined action of MARKs and opposing phosphatases dynamically monitor microtubule assembly state (Figure 4B). One output of the MARK-phosphatase pathway is homeostatic feedback regulation of microtubule stability via MAP phosphorylation, but other outputs are likely (figure 4B). A key missing piece in this model is the mechanism by which microtubules positively regulate phosphorylation of MARK substrates. MARKs bind to microtubules in cells, but also to membranes, actin and adhesion complexes⁶⁴. Adding microtubules to a biochemical assay slowed Tau phosphorylation by a MARK⁵⁸ suggesting that MAPs are phosphorylated when dissociated from the lattice. Perhaps MARKs autoactivate when bound to microtubules, like Aurora-B kinase⁴¹. The transducers in figures 4A and 4B do not act in isolation. MARK2 binds to GEF-H1 and phosphorylates it, causing it to dissociate from microtubules and inhibiting its GEF activity⁶⁵,⁶⁶. Since GEF-H1 activity increases and MARK phosphorylation decreases when microtubules are destabilized, the two transducers likely cooperate to regulate downstream pathways.

A general question for microtubule-transducer pathways is the quantitative nature of their input-output functions. We need to know the degree of change in polymerization dynamics or lattice structure needed to activate a transducer and the fraction of microtubules that must change for the resulting signal to become strong enough to influence a local region or the whole cell. These parameters will determine whether a single microtubule can signal, or rather an entire array must change. A FRET biosensor based on GEF-H1 revealed local activation of RhoA signaling at the leading edge of motile cells in regions where single microtubules recently depolymerized, with a time lag of ~10sec, implying single microtubules are sufficient to signal locally⁴⁹. Low concentrations of nocodazole which destabilized dynamic microtubules in the periphery but left axonal microtubules intact were sufficient to activate protective gene expression in cultured neurons⁴⁶. These observations are consistent with the model in figure 2B where disassembly of a subset of microtubules is sufficient to trigger global signaling. In neurons, damage signals translocate from the periphery to the nucleus by dynein-mediated transport along axonal microtubules⁶⁷,⁶⁸.

Outputs - signaling

Multiple signaling pathways have been implicated in the response to microtubule perturbation in addition to those shown in figures 4A and 4B. Drugs that stabilize and destabilize microtubules both activate mitogen-activated protein kinase (MAPK) signaling, though the destabilization response is typically stronger⁶⁹. In leukocytes and hematological cancer cells, microtubule depolymerization activates MAP3K1 (MEKK) leading to JNK activation⁷⁰,⁷¹. In neurons, partial microtubule depolymerization is sufficient to activate dual leucine zipper kinase (DLK, MAP3K12), also leading to JNK activation⁴⁶. MAP3K1 and DLK are not thought to bind directly to microtubules, though MAP3K1 contains a TOG tubulin binding domain of unknown function⁷². Rather, they are likely activated indirectly via microtubule-binding transducers such as GEF-H1. Precisely how microtubule destabilization activates MAPK pathways in diverse cell types is an important and still unanswered question. Acute MAPK activation helps cells adapt to stress and survive in the short term, but prolonged activation can promote cell death. Prolonged activation of MAP3K1-JNK is thought to mediate the cytotoxic action of vinca alkaloids in hematological cancers⁷⁰,⁷¹ and prolonged activation of DLK has been implicated in neurodegeneration⁷³. This temporal distinction between protective and cytotoxic activation of the same signaling pathways informs design criteria for new microtubule drugs, as discussed below.

Outputs - Gene Expression and Cell Physiology

Most information comes from drug-induced perturbation of microtubules. If we assume drug responses represent exaggerated versions of physiological responses, we can query them to infer how cells respond and adapt to microtubule loss and stabilization. Both perturbations have large effects on gene expression, but as with signaling pathways, depolymerization tends to be the stronger stimulus⁵³.

Transcriptome analysis of microtubule drug responses have been performed in multiple cell types. As an illustration of this type of analysis, figure 5 shows an example of significant changes in gene expression 6hrs after depolymerizing microtubules in two cell types in culture. For clarity, only changes common to both cell types are shown. The physiological implications of gene expression analysis are most informative when the data are collected from animal tissues or primary cells. In rat liver, colchicine induced pro-inflammatory cytokines and drug metabolism enzymes as well as secreted factors which mediate its systemic anti-inflammatory action, including a form of GDF15¹⁶. My interpretation of these data is that colchicine induces inflammation in the liver as a response to tissue damage, in common with other liver other toxins. The secreted anti-inflammatory factors may serve to reduce leukocyte extravasation in non-affected tissues and thus maximize leukocyte availability at the damaged tissue. The transcription factor NRF2 was required for the gene expression response to colchicine in hepatocytes¹⁶. Colchicine was hepatotoxic in NRF2^−/− mice, but not wt., showing that colchicine-induced gene expression is cytoprotective as well as pro-inflammatory. In dendritic cells, microtubule depolymerization induced pro-inflammatory and immune regulatory genes which promoted maturation and anti-tumor immunity⁵². In cultured neurons, mild destabilization of microtubules induced stress, growth and actin cytoskeleton genes while downregulating mature neuron genes, which was interpreted as a cell repair response⁴⁶. In primary endothelial cells, microtubule depolymerization by three different drugs induced similar responses which included induction of genes that promote communication with the immune system, growth factors and blood clotting and were interpreted as a tissue repair response⁵³. Summarizing across studies, destabilizing microtubules rapidly induces gene expression programs that that promote cell survival, growth, inflammation and repair.

Figure 5. Examples of gene expression changes 6hrs after depolymerizing microtubules.

Each dot is a gene whose expression changed significantly in RNAseq data collected 6hrs after drug induced microtubule depolymerization in human pulmonary microvascular endothelial cells (HPMEC, X axis) and quiescent RPE1 cells (Y axis). Growth factor genes whose upregulation is likely to trigger cell proliferation are shown as red dots (PDGFB = Platelet derived growth factor B, HBEGF = Heparin binding EGF like growth factor). For details see refs⁵³,⁹¹.

The pro-growth effect of destabilizing microtubules was first noted in classic studies showing that brief microtubule depolymerization triggers quiescent fibroblasts to re-enter the cell cycle⁷⁴ while stabilization inhibits cell cycle re-entry promoted by growth factors⁴⁰. In a dramatic recent example, transfection of mouse liver with the microtubule destabilizing protein stathmin-1 caused a transient increase in hepatocyte number and doubling of liver mass⁷⁵. Stathmin-1 is commonly overexpressed in cancer⁷⁶, perhaps because it can promote growth via microtubule depolymerization⁷⁵ but is inactivated by phosphorylation during mitosis⁷⁷. The mechanistic basis of the pro-growth effect of destabilizing microtubules is not known, but the genes induced by microtubule disassembly in multiple cell types include growth factors such as HBEGF, PDGF and LIF (shown boxed in figure 5). Induction of these genes is likely to stimulate growth through juxtracrine and paracrine signaling. The observation that microtubule destabilizing drugs which are used to treat cancer, such as vinca alkaloids, have the potential to induce cell cycle re-entry in quiescent cells is somewhat paradoxical. Careful experiments with primary leukemia cells showed that vincristine kills G2/M cells via mitotic perturbation and G1 cells by some alternative mechanism that involves signaling pathway activation, with the G1 pathway more important for therapy⁷⁸.

Gene expression responses to microtubule destabilization also promote adaptative stabilization of microtubules. The best characterized example in animals is induction of a cold-stabilizing tubulin isotype by cold in Drosophila³⁰. Presumably, the complex gene expression profiles induced by destabilization in mammalian cells include such adaptive changes, but these have not been teased out from more general stress, repair and growth pathways.

Design principles for a cytoprotective microtubule-destabilizing drug

The signaling and gene expression programs induced by destabilization of microtubules can be therapeutic, as manifested by the anti-inflammatory activity of colchicine²⁵ and the bone marrow protective activity of plinabulin⁷⁹. However, microtubules are essential for mitosis and neuronal homeostasis, and poisoning these processes causes dangerous side effects when microtubule drugs are used at high doses for cancer therapy. The net clinical benefits of colchicine and plinabulin were discovered empirically, colchicine as a traditional medicine⁸⁰ and plinabulin as an unexpected finding in anti-cancer trials⁷⁹. The sensor hypothesis suggests criteria for rational design of safer and more effective alternatives.

Cell type selectivity:

Targeting specific cell types can dramatically improve therapeutic index. Colchicine is remarkably selective for hepatocytes in mice¹⁶. We speculated that this is due to selective uptake via hepatocyte-specific drug transporters. Alternatively, cell type selectivity might be achieved by targeting specific tubulin isotypes, which is an active research area⁸¹,⁸².

Dose:

Complete microtubule depolymerization is toxic in multiple cell types. It is therefore important to develop dosing regimens which elicit safe degrees of depolymerization in both target and non-target cell types. Colchicine is a good example, where 0.6mg/day is therapeutic while a single dose of 40mg or more is lethal²⁵.

Association-dissociation rates:

Tuning these parameters dramatically affects pharmacokinetics and tissue distribution. Colchicine associates and dissociates extremely slowly from tubulin⁸³. Its slow association retards tubulin engagement in most tissues and helps promote liver-selective distribution (see PK-PD model in¹⁶). The colchicine site ligand plinabulin dissociates much faster⁵³ and correspondingly fast association kinetics likely contribute to its ability to distribute to the bone marrow where it promotes white blood cell production⁷⁹.

Pulsatile pharmacodynamics:

Inducing beneficial signaling and gene expression only requires brief destabilization of microtubules while anti-mitotic toxicity requires that the bone marrow be exposed continuously for the duration of the cell cycle. Furthermore, pulsatile activation of MAPK signaling favors cryoprotection while continuous activation promotes cell death. The ideal drug would therefore dissociate and be eliminated rapidly to generate a sharp pulse of signaling. Modest affinity and rapid metabolism/excretion are therefore important design criteria, combined with dosing regimens that optimize the duration of exposure.

Sub-stochiometric activity:

Colchicine potently inhibits microtube polymerization because only a small fraction of tubulin needs to bind drug for strong inhibition of polymerization⁸⁴,⁸⁵. Given the need for fast dissociation from a safety perspective, and therefore modest binding affinity, sub-stochiometric activity is an important design criterion.

Indications:

In theory, many diseases could benefit from a drug that promotes cell survival, repair and growth, but identification of responsive indications where the benefit/risk ratio of a novel microtubule drug is favorable will be challenging. Traumatic, inflammatory and degenerative disease of the nervous system are all areas of high unmet need. Neurons are highly sensitive to induction of protective gene expression⁴⁶ suggesting that mild, transient microtubule destabilization might be beneficial. Perhaps neural regeneration, which only requires transient drug treatment, is a more promising indication than neurodegeneration, where treatment would need to be life-long. Microtubule stabilization and destabilization both provided benefit in animal models of spinal cord injury⁸⁶,⁸⁷. In my view, more effort should be invested in the destabilization strategy using tool compounds optimized for CNS penetration.

Conclusion and action items

The effector functions of microtubules as structural elements, force generators and transport tracks are now well understood. In comparison, their roles as signal initiators and modulators are fertile ground for new discovery. Based on the data and arguments discussed above, I propose that sensory roles of dynamic microtubules are as or more important than architectural roles in cells with stable tissue morphologies. In cell types where microtubules play essential architectural roles, such as neurons, their shared signal-regulating functions are critical for homeostasis and response to damage. In cells where the architectural roles of microtubules are largely dispensable, such as gut enterocytes⁸⁸, modulation of signaling pathways may be their most important function. This is a radical re-think of microtubule biology which opens exciting new directions for fundamental inquiry and drug development.

Action items for microtubule biologists include measurement of changes in microtubule dynamics and downstream signaling in models of tissue damage and disease, characterization of molecular mechanism and input-output relationships for GEF-H1 and the MARKs and discovery of additional transducers of polymerization dynamics and lattice structure. In plants, a minus end directed kinesin translocates from microtubules to the nucleus in response to cold-induced depolymerization, where it triggers adaptive gene expression⁸⁹ showing that motors can transduce signals from microtubule perturbation. Transducers that respond to microtubule stabilization or/and taxane-induced lattice changes are of particular interest for understanding how taxanes cause solid tumor regression, a research direction that could lead to less toxic replacement drugs. Paclitaxel was reported to cause dissociation of CRMP2 (Collapsin Response Mediator 2, DPYSL2) from mitotic microtubules, pointing to one candidate transducer⁹⁰.

For medicinal chemists, my proposed action item is development of microtubule destabilizing tool compounds optimized for non-toxic destabilization of microtubules according to the criteria discussed above. These criteria are very different from the traditional emphasis on maximizing binding affinity and cytotoxicity to proliferating cells²¹ and might elicit a renaissance in colchicine site medicinal chemistry. Colchicine keeps finding new medical uses despite its lack of rational optimization⁸⁰. Improved microtubule modulating drugs are conceptually and chemically feasible and might be useful in a range of diseases where transient stimulation of cell survival, repair and growth pathways would be therapeutic.