Müller-Deku et al. 2019, AzTax | 1

Photoswitchable microtubule stabilisers optically control tubulin

cytoskeleton structure and function

Adrian Müller-Deku1, Kristina Loy1, Yvonne Kraus1, Constanze Heise1, Rebekkah Bingham1, Julia

Ahlfeld1, Dirk Trauner2,*, Oliver Thorn-Seshold1,*,#

1: Department of Pharmacy – Center for Drug Research, Ludwig-Maximilians University, Butenandtstrasse 5-13, Mu-nich 81377, Germany; 2: Department of Chemistry, New York University, 100 Washington Square East, New York NY 10003, United States of America. Keywords: microtubule stabiliser, tubulin polymerisation inhibitor, photopharmacology, taxol, cytoskeleton.

ORCIDs: J.A. 0000-0002-4879-4159; D.T. 0000-0002-6782-6056; O.T.-S. 0000-0003-3981-651X * Senior Authors # Correspondence and requests for materials to O.T.-S. (oliver.thorn-seshold@cup.lmu.de)

ABSTRACT: Small molecule inhibitors provide a versatile method for studies in microtubule cytoskeleton research, since tubulin is not readily amenable to func-tional control using genetics. However, traditional chemical inhibitors do not al-low spatiotemporally precise applications on the length and time scales appro-priate for selectively modulating microtubule-dependent processes. We have syn-thesised a panel of taxane-based light-responsive microtubule stabilisers, whose tubulin hyperpolymerisation activity can be induced by photoisomerisation to their thermodynamically metastable state. These reagents can be isomerised in live cells, optically controlling microtubule network integrity, cell cycle repartition, and cell survival, and offering biological response on the timescale of seconds and spatial precision to the level of individual cells. These azobenzene-based microtubule stabilisers offer the possibility of noninvasive, highly spatiotemporally precise modulation of the microtubule cytoskeleton in live cells, and can prove powerful reagents for studies of intracellular transport, cell motility, and neurodegeneration.

ON

360 nm

OFFdark

OH O OH

OOAc

H

O

OHPh

NH

OHO

OBz

NN

R2

R1

R2

O

R3

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 2

Introduction The cellular cytoskeleton, built from F-actin,

microtubules and intermediate filaments, is a scaffold for critical biological processes from signaling and cargo trafficking, to cell shape maintenance and cell di-vision. Most of the cytoskeleton's myriad biological roles are inherently spatially and temporally differenti-ated, although all rely on the same three protein scaf-fold structures. Aiming to study these cytoskeleton-de-pendent processes with the spatiotemporal resolution appropriate to their biological function, a range of re-cent approaches have used optogenetic methods to photocontrol selected cytoskeleton-associated pro-teins, so enabling their spatiotemporally precise ma-nipulation.1–3 More broadly however, direct spatiotem-poral control over the structure and dynamics of the cy-toskeleton scaffold proteins themselves, would offer a general approach to modulate any of their structure-de-pendent functions - although no genetic methods have yet been shown to deliver this.

We here focus on the microtubule (MT) cyto-skeleton. MTs play particularly important roles in in-tracellular transport, cell motility, and structural plas-ticity, and there is an eminent need to achieve a better understanding of how these many functions are imple-mented and regulated.4 The role of MT dynamics dur-ing cell proliferation has also made them a major anti-cancer target, for which several outstanding drugs (tax-anes, epothilones, vinca alkaloids) have been devel-oped.5–7 In biological research, these drugs and other small molecule modulators (e.g. nocodazole, com-bretastatin, peloruside) remain the most general tools for MT cytoskeleton research. However, these inhibi-tors suppress all contemporaneous MT-dependent functions spatially indiscriminately: so they too do not allow spatiotemporally precise applications on the length and time scales appropriate for selectively stud-ying MT-dependent processes. This restricts their scope of applications and their utility for selective re-search into MT cytoskeleton biology.8

Deeper insights could be gained from inhibitors that allow spatiotemporally specific MT manipulation. Photopharmaceuticals – photoswitchable (exogenous) small molecule inhibitors that act as an optically-con-trolled interface between a researcher and a protein of interest – have developed greatly in recent years.9,10 Photopharmaceuticals conceptually enable studies not otherwise accessible to biology, marrying the spatio-temporal precision of light application known from

optogenetics, to the flexibility and system-independ-ence of exogenous small molecule inhibitors, in a way that particularly favours noninvasive studies of tempo-rally-regulated, spatially-anisotropic biological sys-tems, such as the MT cytoskeleton.9,11,12 Photopharma-ceuticals have succeeded in delivering a measure of op-tical control over a broad range of biochemical and bi-ological phenomena, with early cell-free studies now supplanted by applications in cellulo and recently in vivo.13–16

In the cytoskeleton field, photopharmaceutical analogues of the MT destabiliser colchicine were re-cently developed, to begin addressing the need for spa-tiotemporally precise MT cytoskeleton studies. The az-obenzene-based Photostatins (PSTs), which can be re-versibly photoswitched by low-intensity visible light between their biologically inactive E-isomers and their MT-inhibiting, colchicine-like Z isomers, were first re-ported in 2014.9,17–19 MT destabilising photopharma-ceuticals based on two different families of molecular photoswitch – styrylbenzothiazoles (SBTubs)20 and hemithioindigos (bi-active HOTubs21 and dark-active HITubs)22 – have since been developed, delivering in-creased metabolic robustness in the intracellular envi-ronment as well as alternative optical switching pro-files (all-visible switching with hemithioindigos, and GFP-orthogonal switching with SBTubs). All three re-agent families have enabled highly spatiotemporally precise optical control over endogenous MT network integrity, MT polymerisation dynamics, cell division and cell death; and the PSTs have already been used in animals to resolve outstanding questions e.g. in mam-malian development13,14 and neuroscience23. These ap-plications illustrate the power of photopharmacology to enable previously inaccessible studies of spatiotem-porally anisotropic cytoskeletal processes without ge-netic engineering.12–14,24

With the optically precise blockage of MT polymerisation addressable by a range of photophar-maceuticals, we now desired to develop photopharma-ceutical MT stabilisers (hyperpolymerisers) as concep-tually novel tools with an alternative spectrum of bio-logical research applications. The biological functions of the MT cytoskeleton are primarily driven by the lo-calisation of stable or growing MTs themselves. There-fore, achieving optically-specific promotion of MT network integrity should allow spatiotemporally pre-cise stimulation of MT-dependent functions, rather than localised blockade.

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 3

Such reagents could be applied to many biolog-ical questions, including further exploring recent dis-coveries of important roles of MTs in developing and regenerating neurons. While during development, MT stabilization has been shown to determine axonal iden-tity and remodeling, MT stabilization in mature neu-rons seems to promote axonal regeneration by reducing the formation of retraction bulbs and modulating glial scar formation after spinal cord injury.25–29 However, the temporal characteristics of these phenomena are unclear, and the roles of MT stabilization in surround-ing glia and immune cells rather than the damaged neu-rons themselves have not yet been resolved. Tempo-rally-specific studies of the roles of dynamic and stabi-lised MTs would be highly desirable for studies in a range of other fields, such as in mitotic progression and immune cell response, particularly if stabilisation can be spatially targeted to selected cells within complex environments. Photopharmaceutical stabilisers could be used to trigger and study these phenomena with high temporal resolution and cell-level specificity, and would promise to shed new light on such roles of MTs in development, neuronal regeneration and repair.25,30

As far as we are aware, Liu and coworkers31 re-cently disclosed the only work in the direction of pho-topharmaceutical MT stabilisation, applying the pho-toswitchability of host-guest interactions between beta-cyclodextran and arylazopyrazoles (up to binding constant 2300 M-1)32 to aim at photoswitchably re-versible, noncovalent heterodimerisation of two paclitaxel conjugates that should crosslink two MTs. The authors report that the heterodimeriser isomer-de-pendently alters the proportion of subG1-phase (dying) cells from 8% to 12% (cis and trans respectively, ± 2%) although in cellulo photoswitchability was not demon-strated. We considered that this work does not solve the need for a robust optical MT reagent, and features con-ceptual drawbacks. Crosslinking by the heterodimer (maximum taxol-to-taxol distance ca. 2 nm) of two MTs seems geometrically unlikely since taxol's bind-ing site is on the luminal (inner) face of the MTs33, giv-ing a minimum taxol-to-taxol distance ca. 7 nm and re-quiring the linker to penetrate directly through both protein walls, as well as stretching substantially and re-orienting the usual taxol binding pose. The lack of cel-lular potency of each monomer half is also striking (perhaps since the critical7 2'-hydroxyl is capped with the photoswitch albeit in a way that may be enzymati-cally labile), and the additional possibility of isomer-

dependent transmembrane trafficking of this ~4 kDa construct complicates results. Such considerations (see discussion in Supporting Information) encouraged us instead to pursue lower-molecular-weight, druglike re-agents that could offer robust, structurally rationalisa-ble performance. We therefore chose to develop mon-omeric paclitaxel analogues incorporating azoben-zenes that directly give photoswitchability of tubulin-binding potency, as optically controlled MT stabilisers for in situ spatiotemporal photocontrol of MT network structure and function. We now report our development of these reagents.

Results Design and Synthesis We chose the azobenzene photoswitch for in-

stalling photoswitchable binding potency to the taxane core. This photoswitch offers a substantial geometric change upon isomerisation, which we hoped would dif-ferentiate the isomers' binding constants, and allows re-liable, high-quantum-yield, near-UV/visible-light-mediated, highly robust E↔Z photoisomerisability, which allows repeated in cellulo photoswitching. Tax-anes feature a number of chemically modifiable posi-tions; we chose to focus on sites where substituents can be tolerated, but where their geometric changes might impact binding potency through steric interactions. Po-tent taxanes feature a side-chain 3'-amine substituted with mid-size hydrophobic groups (e.g. Boc group in docetaxel, Bz in paclitaxel)7,34 which abut the tubulin protein surface yet are projected away from the protein interior (Fig 1a, highlighted in pink); the other sidechain positions (eg. the 3'-phenyl or 2'-hydroxyl) offer less spatial tolerance for substitution as they pro-ject into the protein7. The 3'-amine also tolerates the attachment of polar cargos such as the large silarhoda-mine fluorophore, as long as they are attached via a long spacer,35 making it desirable for photopharmaceu-tical tuning as it might tolerate azobenzenes with a range of structural characteristics.

We accordingly designed a panel of 3'-azoben-zamide-taxanes (AzTaxs) for biological testing. As taxanes have famously poor aqueous solubility (still worsened by attaching an azobenzene), we initially de-termined to focus on compounds displaying satisfac-tory potency at concentrations substantially below their solubility limit. This avoids the case that the com-pounds' apparent potencies would be dictated by solu-

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 4

bility effects, and so should enable robust use as rea-gents across a variety of systems and settings. Theoris-ing that the sterics around the azobenzene phenyl ring proximal to the taxane core would be the greatest po-tency-affecting factor, we first focussed on testing which orientions of photoswitch would be best toler-ated. We therefore scanned orientations of the diazene in ortho, meta and para relative to the amide (Az-Tax2/3/4 compound sets), and when early testing showed that the AzTax2 set had the lowest potency, we abandoned it at this stage. Next, examination of the published tubulin:paclitaxel cryo-EM structures33,36 in-dicated that the azobenzene's distal ring can project freely away from the protein. Therefore we hypothe-sised that steric variation to the distal ring would not greatly impact binding potency of either isomer, but could be used orthogonally to tune their photochemical properties, by substitutions in para to the diazene that can mesomerically affect the properties of the N=N double bond. We accordingly synthesised unsubsti-tuted ("H"), para-methoxy ("MP"), and para-dime-thylamino ("DMA") derivatives of the AzTax3/4 sets. These were chosen to vary the photochemical proper-ties of most relevance to photopharmacology: the com-pleteness of the E→Z and the Z→E photoisomerisa-tions at fixed wavelengths, which dictate the dynamic range of isomer photoswitchability, and τ (the halflife of the spontaneous unidirectional Z→E relaxation). Lastly, when the AzTax3 set proved promising in early studies, we also examined installing an electron-donat-ing 3,4,5-trimethoxy motif on the distal ring (Az-Tax3TM) as well as an additional R3 methoxy group to reduce the rotatability of the proximal ring in case this could amplify the difference between isomer po-tencies (AzTax3MTM), and we controlled for solubil-ity effects by exchanging the dimethylamino substitu-ent for a more soluble diethanolamino (AzTax3DEA). The target AzTaxs were synthesised by degradation of commercial docetaxel followed by amide couplings to various azobenzenecarboxylic acids in moderate yields (Fig 1b-c).

Figure 1 Design and synthesis of AzTaxs. (a) Paclitaxel:tu-bulin structure (PDB: 3J6G33) with the benzamide indicated in pink. (b) Synthesis of AzTaxs from docetaxel. (c) Panel of AzTaxs examined in this study.

Photochemical characterisation The AzTaxs all displayed robust and repeatable

E↔Z photoswitching under near-UV/visible illumina-tions, as expected from the literature11 (Fig S1). The photochemical properties within each substituent set were similar. The "H" compounds displayed a 3-fold dynamic range of Z-isomer photoswitchability be-tween the photostationary states (PSSs) at 375 nm (80% Z) and 410 nm (26% Z), and had substantially slower relaxation than biological timescales (τ ca. 50 days). The methoxylated compounds ("MP", "TM" and "MTM") had been chosen to improve the dynamic

OR’ OMe

OH

OOAc

H

O

OHPh

NHR

OHO

OBzEDCIHOBt DIPEA

OH OMe

OH

OOAc

H

O

OHPh

NH

OHO

OBz

NN

R2

R1

R2

O

CO2H

Azo R = Boc: docetaxel (R’ = H)R = Bz: paclitaxel (a; R’ = Ac)

1. TFAdocetaxel:

b

a

c

R3

AzTaxseries

o/m/p

2.

compound attach R1 R2 R3

AzTax4H para H H HAzTax3H meta H H HAzTax4MP para OMe H HAzTax3MP meta OMe H HAzTax2MP ortho OMe H HAzTax3TM meta OMe OMe HAzTax3MTM meta OMe OMe OMeAzTax4DMA para NMe2 H HAzTax3DMA meta NMe2 H HAzTax3DEA meta N(C2H4OH)2 H H

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 5

range of photoswitching by relative shifting of the iso-mers' absorption bands9 (Fig S2), and delivered a ca. 9-fold dynamic range of Z-isomer photoswitchability (375 nm: 96% Z; 530 nm: 11% Z); their relaxation re-mained substantially slower than biological timescales (τ ca. 24 h). The para-amino "DMA" and "DEA" com-pounds featured τ values too small to observe bulk pho-toswitching in aqueous media under biologically appli-cable conditions. Yet, as aprotic environments (such as lipid vesicles, membranes, and on-protein adsorbed states) are likely intracellular localisations for hydro-phobic taxanes conjugates, we determined their photo-chemistry in moderately polar aprotic media (EtOAc). Here they were easily bulk-switchable (τ ca. 11 min), giving a 4-fold dynamic range of Z-isomer pho-toswitchability (410 nm: 91% Z; 530 nm: 21% Z) (fur-ther discussion in the Supporting Information). As the AzTax reagents were intended for use with micros-copy, we also examined photoswitching over a broader range of wavelengths, to determine what dynamic range of isomer photoswitchability would be accessi-ble in practice, with standard (405 nm, 488 nm, 514 nm) or more exotic (380 nm, 440 nm, 532 nm) mi-croscopy laser wavelengths (Fig 2a, Fig S2, Table S1).

We next proceeded to explore the biological ap-plicability of AzTaxs as photoswitchable MT stabilis-ers in cellulo. Since near-UV light gave PSSs with high-Z populations for all photoswitches, while ther-mal relaxation and maintenance in the dark returned the E isomer quantitatively, we began by comparing all-E "dark" (all-E stock applied, then maintained dark) with mostly-Z "360 nm" (all-E stock applied, then pho-toisomerised in situ by pulsed illuminations with low-power 360 nm LED light, giving a mostly-Z PSS) con-ditions, to determine which structures allowed the highest fold difference of bioactivity.

AzTaxs display optically-controlled bioactiv-ity in cellulo

Since stabilisation and hyperpolymerisation of MTs in cells over a prolonged period blocks cell pro-liferation and ultimately causes cell death7, we first as-sayed the AzTaxs for light-dependent cellular activity by the resazurin cell proliferation/viability assay. Via-bility dose-response curves under dark or UV condi-tions were assessed in the HeLa cervical cancer cell line (Fig 2b-c). All compounds displayed dose-re-sponse curves with similar Hill coefficients as the par-ent drug docetaxel (Fig S3), which is in line with the conjecture that they act through the same mechanism,

albeit with different potency. All compounds except the fast-relaxing AzTax3DMA had Z-isomers that were more potent, or else equipotent, to the E isomers, suggesting that this trend in isomer-dependent cellular bioactivity across several photoswitch types has robust significance. E-AzTax2MP had the poorest overall po-tency (EC50 ca. 7 µM, all-E), which we took as indicat-ing the unsuitability of ortho substitutions that likely project the azobenzene into solution (c.f. Fig 1a) where its hydrophobicity can defavour binding stability. By contrast, the AzTax4 set featured compounds up to 100 times more potent, and structure-dependently cov-ered a 40-fold potency range. However, despite the good isomeric photoswitchability of e.g. AzTax4MP, none of the AzTax4 set displayed substantial pho-toswitchability of bioactivity (fold difference between the 360 nm and the dark bioactivity). We interpreted this substitution-independent result as an indication that the distal ring may project too far from the protein contact surface, in both E and Z isomers, for the Az-Tax4 isomer state to substantially affect binding.

In contrast, the AzTax3 series all showed pho-toswitchability of bioactivity. AzTax3MP featured a nearly 6-fold difference between the more-toxic Z and less-toxic E isomers' bioactivity, over the 48 h experi-mental timecourse (Fig 2b). Increasing the number of methoxy groups on the scaffold slightly decreased the Z-isomer's cytotoxicity without greatly affecting that of the E-isomer (AzTax3TM, AzTax3MTM), and delet-ing the methoxy group also decreased the Z-isomer's cytotoxicity (AzTax3H), which we took as a sign that balancing the photoswitch's polarity was important for maximising bioactivity. In line with this interpretation, the potencies of AzTax3DMA were similar to Az-Tax4DMA while the more hydrophilic AzTax3DEA showed a 40-fold loss of potency. Despite the potential for photoswitching the para-amino AzTax inside lipid environments, they should only reach their cytosolic target tubulin as the E-isomers due to fast aqueous re-laxation, yet surprisingly, AzTax3DMA appeared slightly more bioactive as the unilluminated E isomer, although as expected AzTax4DMA and AzTax3DEA both showed no illumination-dependency of bioactiv-ity; and controls under 410 nm illumination (to estab-lish the optimum PSS for the para-amino compounds) showed no different result to those obtained with UV illumination. The apparent cytotoxicity differential seen for AzTax3DMA might reflect reduced availabil-ity to the cytosol rather than differential binding of the

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 6

isomers22 although these results do not allow further conjecture.

To continue the study we therefore selected Az-Tax3MP, due to its photoswitchability of bioactivity (6-fold), high potency (0.24 µM when UV-illuminated), bidirectional photoswitchability (opti-mum 9-fold-change of concentration of the more bio-active Z isomer), and reproducibly photoswitchable cellular performance across assays with different illu-mination conditions, and proceeded with further mech-anistic biological evaluations.

Figure 2 Photoswitchable performance of AzTaxs. (a) Pho-tostationary state UV-Vis absorption spectra of AzTax3MP under a range of cell-compatible wavelengths similar to mi-croscopy laser lines. (b-c) Resazurin antiproliferation assays of AzTaxs highlight their structure- and light-dependent cell cytotoxicity. HeLa cells, 40 h incubation in dark conditions (all-E) or under pulsed illuminations with low-power LEDs (75 ms per 15 s near-UV at <1 mW/cm2; “lit” = ~80% Z). (d) A cell-free assay for polymerisation of purified tubulin com-paring the MT stabilisation activity of docetaxel, all-E- and 360 nm-lit-AzTax3MP (all 10 µM) shows light-specific in-duction of hyperpolymerisation by Z-AzTax3MP, matching the trend observed in cellular assays.

AzTaxs photocontrol tubulin polymerisation, cellular MT networks and cell cycle

To examine the molecular mechanism of AzTax isomer-dependent cellular bioactivity, we first assayed the potency of AzTax3MP for tubulin hyperpolymeri-sation in cell-free assays using purified tubulin. The majority-Z 360 nm-lit state gave a ca. 60% enhance-ment of polymerisation over control (benchmarked to

docetaxel at 100%), while all-E-AzTax3MP gave only ca. 30% polymerisation enhancement (Fig 2d). This clarifies the mechanism of action of AzTaxs as MT sta-bilisers, like their parent taxanes.

We next studied the direct effects of in situ-pho-toisomerised AzTax upon cellular MT structure and MT-dependent processes. Immunofluorescence imag-ing in cellulo revealed that AzTax3MP causes light-dependent disruption of the MT network, resulting in mitotically-arrested cells, mitotic spindle defects, and multinucleated cells as its concentration increases (Fig 3a). The best window for visualising this isomer-de-pendent bioactivity lay around 0.3-1 µM (Fig S5). Z-stack projections additionally revealed cells which were substantially accumulated into rounded, mitoti-cally arrested states that are not well resolved in single-plane imaging (Fig S5). Both the mitotic arrests, and the nuclear defects of cells that escape arrest, are hall-marks of MT stabiliser treatment37, arguing that the isomer-dependent cytotoxicity of AzTax3MP in cel-lulo arises from MT stabilisation preferentially by its Z-isomer.

The multinucleated cells indicated that AzTax also inhibits MT-dependent functions such as success-ful completion of mitosis. To quantify this we exam-ined cell cycle repartition after AzTax treatment by flow cytometry, expecting to observe G2/M-phase-cell cycle arrest.38 G2/M-arrest was observed with approxi-mate EC50 around 1.5 µM for the "lit" AzTax3MP, twice as potent as E-AzTax (Fig 3b-d), mimicking the effect of docetaxel although with lower potency (Fig S4). As a control for illumination/photoswitch-depend-ent off-target effects, we examined the non-pho-toswitchably-bioactive but potent AzTax4DMA, which reproduced the effects of docetaxel independent of illumination conditions indicating no significant as-say complications (Fig S4).

This further supported that AzTax3MP acts across a range of assays and readouts as a light-modu-lated taxane, with reproducible photocontrol over the isomers' bioactivity - both directly against MTs, and in-directly against MT-dependent processes.

AzTax3MPPSS equilibria

AzTax3MPcell viability

tubulinpolymerisation

-7 -6 -50

50

100

log10([AzTax3MP]) (M)

cell

viab

ility

(%) dark

360 nm

a b

300 400 5000.00

0.25

0.50

0.75

wavelength (nm)

Abs

dark530 nm515 nm490 nm435 nm410 nm375 nm

dEC50 ratiocompound 360nm dark dark / litAzTax4H 0.048 0.082 1.7AzTax3H 0.64 0.89 1.4AzTax4MP 2.1 2.0 1.0AzTax3MP 0.24 1.4 5.8AzTax2MP 1.8 6.8 3.7AzTax3TM 0.53 1.6 3.0AzTax3MTM 0.65 1.2 1.9AzTax4DMA 0.13 0.15 1.1AzTax3DMA 0.20 0.096 0.5AzTax3DEA 4.3 4.7 1.1

EC50 (µM)c

0 1 2 3

0.1

0.5

time (min)

Abs

340

cosolvent

docetaxel

E-AzTax3MPlit AzTax3MP

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 7

Figure 3 AzTax3MP disrupts MT structure and MT-dependent functions in cellulo. (a) Immunofluorescence staining for MT structure shows dose- and light-specific MT disruption (see also Fig S5). HeLa cells incubated for 20 h; α-tubulin in green, DNA in blue, scale bars 20 µm. (b-d) Flow cytometry analysis of cell cycle repartition shows that AzTax3MP dose-dependently gives G2/M (see also Fig S4) similar to that reached with positive control docetaxel (0.1 µM).

Discussion Photocontrol over protein function represents an

attractive method to study anisotropic, multifunctional cellular systems, as it offers to address complex biol-ogy with the spatiotemporal specificity required to fo-cus on specific roles or aspects of the system. Small molecule photopharmaceuticals have already proven valuable for their unique ability to address such targets that are not directly accessible to optogenetics, such as the MT cytoskeleton, for which a range of pho-toswitchable depolymerising agents have recently been reported9,20,21. Here we have expanded the scope of photopharmaceutical cytoskeleton reagents to demon-strate the first photoswitchable MT hyperpolymerising agents. Through early structure-photochemistry/activ-ity-relationship studies, we have identified a lead com-pound AzTax3MP that gives robust, in situ-pho-toswitchable MT stabilising activity in cell-free and cellular assays, and can light-dependently reproduce

key direct as well as downstream biological effects of the taxanes. This is a promising starting point for fur-ther reagent optimisation, and we believe that Az-Tax3MP itself will already find a range of applications particularly in embryology, neuroscience and motility, where its spatiotemporally-specific bioactivity will en-able studies not previously possible.

Determining the sources of the differential bio-activity between AzTax isomers in cellulo is key for optimisation. Since modifying polarity at a distal site that should not clash sterically with the protein gave a 40-fold change of apparent potency (AzTax3DEA compared to AzTax3DMA), the sterics of the AzTax isomers are not necessarily the sole determinant of cel-lular bioactivity. Yet, polarity-dependent cellular biolo-calisation or penetration cannot entirely explain the photoswitchable activity of AzTax3MP since it shows isomer-dependent activity in cell-free assays also, so

S G2/MG1sub-G1

cosolvent1.5 µM dark1.5 µM 360 nm0.1 µM docetaxel

0

50

100

%PI-p

ositive

cells

AzTax3MP cell cycleAzTax3MP cell cycle

sub-G1

S

G1

G2/M

0 0.1 0.75 1.5 3.0d λ d λ d λ d λλ

C [µM]d -7 -6.25 -5.75

0

20

40

60

80

%ce

llsinG2/M

0.1 µM docetaxel

cosolvent control

360 nm

dark

log10([AzTax3MP]) (M)

G2/M arrestb c d

a

pos.

360nm

dark

0.5 µM controls1 µM 1.5 µM

0.1 µM docetaxel

cosolvent

α-tubulinDAPI

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 8

the azobenzene must significantly impact protein-lig-and affinity. We believe that maximising the bioactivity difference between isomers will require photoswitches with isomer-dependency both of sterics and of polarity. The E→Z photoswitchability of the azobenzene was not correlated to the photoswitchability of biological activity (c.f. AzTax3MTM, AzTax4MP). As the Z-AzTax were typically the more-active isomers, further defavouring the binding of the E-isomer while allow-ing the Z-isomer to retain bioactivity is likely the best way to maximise the photocontrol over inhibition. Re-search in these directions is underway.

Photopharmacology often assumes increasing the dynamic range of isomeric photoswitchability un-der a freely available choice of illumination conditions, and redshifting overall absorption wavelengths, are re-quired for improved biological performance. In the case of AzTax stabilisers, probably neither considera-tion applies. Once MT stabilisation and hyperpolymer-isation is induced by E→Z isomerisation, the altered MT biology probably cannot be instantaneously re-turned to its usual state even if the stabiliser would be totally removed (e.g. by complete optical, or thermal, relaxation to an inactive state). Stabilised MTs that have hyperpolymerised, potentially with loss of direc-tionality, presumably require time to break down and return tubulin monomer to the cytoplasmic pool, so that ordinary MT structures can be rebuilt. Therefore there is probably a limit to the temporal resolution of true biological reversibility that any photoswitchable stabiliser can display, even if selected readouts (such as speed of polymerisation of individual MTs) can re-cover more quickly. With this consideration in mind we do not believe that improving the completeness of bi-directional isomeric photoswitchability39 will be as im-portant for AzTax development as for other classes of inhibitors that can feature instantaneous biological re-sponse. Redshifting photoswitch absorption wave-lengths is also likely to be counterproductive for a mi-croscopy reagent, since there are few fluorescent pro-teins with significant excitation efficiency at laser lines above 561 nm (typically the next wavelength available is 647 nm); maintaining orthogonality to the widest possible range of imaging wavelegnths by blueshifting is probably more advantageous.20 However, a key property that should be readily tunable to the advantage of this system is the E→Z photoisomerisation effi-ciency at 405 nm, which is usually the only microscopy

laser available in the 350-440 nm range. Here we con-sider that improved performance for AzTax-like rea-gents will depend on optimising photoconversion at the wavelength/s that will in practice be used for their pho-tocontrol. Developing a set of standard photoswitches with better 405 nm E→Z photoconversion than these para-alkoxyazobenzenes (~46% Z) yet with similar polarity and substantial stability against thermal relax-ation, is a nontrivial goal of our ongoing research.

Outlook The AzTax photoswitchable microtubule stabi-

lisers can be used in conjuction with long-term, in situ photoswitching in live cells to control fundamental bi-ological processes from cytoskeleton architecture to cell survival. By complementing the existing MT-depolymerising photopharmaceuticals, AzTaxs now bring both principal modes of MT regulation under op-tical control. Through structure-photochemistry/activ-ity-relationship studies we have identified perspectives for improving their biological photocontrol. This opens up several avenues for applying fundamental research in the rapidly evolving field of chemical photoswitches to generate specialty MT stabiliser photopharmaceuti-cals for cell-free mechanistic studies, cell biology, and towards in vivo use. We consider, more broadly, that this work will also encourage further photopharmaceu-tical work on other proteins inaccessible to direct opto-genetics, such as the actin cytoskeleton.

AzTax reagents offer to aid in studying MT bi-ology particularly where the temporally- or cell-type-specific biological roles of MT stabilisation are un-clear, thus addressing a range of biological questions across the fields of neuroscience, development, cell di-vision, signaling and migration.25,30 Spatially-selective induction of MT hyperpolymerisation and structural stabilization will also be of particular interest for stud-ies where localized maintenance or growth of MT net-works is thought to drive biology. Such high-spatio-temporal-precision reagents also offer an intriguing method to study the temporal and spatial dependency of biological action of their “parent” taxanes. To a large degree it is still37 unclear how taxanes exert their cel-lular/tissue-level effects in vivo. Yet there is enormous clinically-driven interest in increasing the understand-ing of taxane pharmacology, both towards improved taxol-site antimitotic therapeutics, and increasingly to-wards designing better combination treatment regimes

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 9

involving these broad-spectrum cancer chemothera-peutics. Thus we believe the AzTaxs will open up a multitude of possibilities for high-precision studies not possible with previous methods, across basic and ap-plied research.

In conclusion, we believe that this first demon-stration of photoswitchable MT stabilisers represents an important step towards high-spatiotemporal-preci-sion studies of critical proteins and processes in cell bi-ology, and that the AzTax will be especially useful in studies of intracellular transport, signaling, cell motil-ity, and neurodegeneration.

Methods in brief Full and detailed experimental protocols can be

found in the Supporting Information. Compound synthesis and characterisation.

Reactions and characterisations were performed by de-fault with non-degassed solvents and reagents (Sigma-Aldrich, TCI Europe, Fisher Scientific), used as ob-tained, under closed air atmosphere without special precautions. Manual flash column chromatography was performed on Merck silica gel Si-60 (40–63 µm). MPLC flash column chromatography was performed on a Biotage Isolera Spektra, using Biotage prepacked silica cartridges. Thin-layer chromatography (TLC) was run on 0.25 mm Merck silica gel plates (60, F-254), with UV light (254 nm and 365 nm) for visu-alization. NMR characterisation was performed on Bruker 400 or 500 MHz spectrometers. HRMS was performed by electron impact (EI) at 70 eV with Thermo Finnigan MAT 95 or Jeol GCmate II spectrom-eters; or by electrospray ionization (ESI) with a Thermo Finnigan LTQ FT Ultra Fourier Transform Ion Cyclotron resonance spectrometer. Analytical HPLC-MS was performed on an Agilent 1100 SL HPLC with H2O:MeCN eluent gradients, a Thermo Scientific Hypersil GOLD™ C18 column (1.9 µm; 3 × 50 mm) maintained at 25°C, detected on an Agilent 1100 series diode array detector and a Bruker Daltonics HCT-Ultra mass spectrometer.

Photocharacterisation. UV-Vis-based studies (determination of absorption spectra, photostationary states, reversibility of photoisomerisation, and Z to E relaxation) were performed on a Varian CaryScan 60 (1 cm pathlength) at room temperature with model photoswitches that were water-soluble analogues of the AzTax species, since reliable UV-Vis studies require compound concentrations around 25-50 µM, while the

AzTax compounds are only reliably molecularly solu-ble at such concentrations with high cosolvent percent-ages (eg. 50% DMSO) that do not reflect the intracel-lular environment and also alter the isomers’ spectra, quantum yields, and relaxation times. We synthesised and used di(2-ethanol)amine carboxamides as water-soluble analogues of the taxane carboxamide AzTaxs (see Supporting Information) enabling measurements in PBS at pH ~7.4 with only 1% of DMSO as cosol-vent, thus matching the intracellular environment around the AzTaxs’ protein target, tubulin. “Star” 3W LEDs (360–530 nm, each FWHM ~25 nm, Roithner Lasertechnik) were used for photoisomerisations in cu-vette that were thus predictive of what would be ob-tained in the cytosol during LED-illuminated cell cul-ture. Spectra of pure E and Z isomers were acquired from the HPLC’s inline Agilent 1100 series diode array detector (DAD) over the range 200–550 nm, manually baselining across each elution peak of interest to cor-rect for eluent composition.

Tubulin Polymerisation in vitro. 99% purity tubulin from porcine brain was obtained from Cyto-skeleton Inc. (cat. #T240) and polymerisation assays run according to manufacturer’s instructions. Tubulin was incubated at 37 °C with “lit”- or “dark”-AzTax (10 µM) in buffer (with 3% DMSO, 10% glycerol) and GTP (1 mM), and the change in absorbance at 340 nm was monitored over 15 mins at 37°C40.

Cell Culture. HeLa cells were maintained un-der standard cell culture conditions in Dulbecco’s mod-ified Eagle’s medium supplemented with 10% fetal calf serum (FCS), 100 U/mL penicillin and 100 U/mL streptomycin, at 37°C in a 5% CO2 atmosphere. Cells were transferred to phenol red free medium prior to as-says. Compounds (in the all-E state) and cosolvent (DMSO; 1% final concentration) were added via a D300e digital dispenser (Tecan). Treated cells were then incubated under “dark” (light-excluded) or “lit” conditions (where pulsed illuminations were applied by multi-LED arrays to create and maintain the wave-length-dependent photostationary state isomer ratios throughout the experiment, as previously described9). “Lit” timing conditions were 75 ms pulses applied every 15 s.

Resazurin Antiproliferation Assay. As a proxy readout for viable cells, mitochondrial diapho-rase activity in HeLa cell line was quantified by meas-uring the reduction of resazurin (7-hydroxy-3H-phe-noxazin-3-one 10-oxide) to resorufin. 5,000 cells/well

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 10

were seeded on 96-well plates. After 24 h, cells were treated with E-AzTaxs, shielded from ambient light with light-proof boxes, and exposed to the appropriate light regimes. Following 48 h of treatment, cells were incubated with 20 µL of 0.15 mg/mL resazurin per well for 3 hours at 37°C. The resorufin fluorescence (exci-tation 544 nm, emission 590 nm) was measured using a FLUOstar Omega microplate reader (BMG Labtech). Results are represented as percent of DMSO-treated control (reading zero was assumed to correspond to zero viable cells) and represented as mean of at least three independent experiments with SD.

Cell Cycle analysis. E-AzTaxs were added to HeLa cells in 6-well plates (seeding density: 300,000 cells/well) and incubated under “dark” or “lit” condi-tions for 24 h. Cells were harvested and fixed in ice-cold 70% ethanol then stained with propidium iodide ("PI", 200 µg/mL in 0.1 % Triton X-100 containing 200µg/mL DNase-free RNase (Thermo Fischer Scien-tific EN0531) for 30 min at 37°C. Following PI stain-ing, cells were analyzed by flow cytometry using an LSR Fortessa (Becton Dickinson) run by BD FACSDiva 8.0.1 software. The cell cycle analysis was subsequently performed using FlowJo-V10 software (Tree Star Inc.). Cells were sorted into sub-G1, G1, S and G2/M phase according to DNA content (PI signal). Quantification from gating on the respective histo-grams is shown as percent of live/singlet/PI-positive parent population per cell cycle phase across different concentrations of the compound. Every experiment was performed in technical triplicates, at least three times independently, with a minimum of 10,000 (mean: 14,000) PI-positive singlet cells analyzed per replicate.

Immunofluorescence Staining. HeLa cells seeded on glass coverslips in 24-well plates (50,000 cells/well) were left to adhere for 24 h then treated for 24 h with AzTaxs under “dark” or “lit” conditions. Cover slips were washed then fixed with 0.5% glutar-aldehyde, quenched with 0.1% NaBH4, blocked with PBS + 10% FCS, treated with rabbit alpha-tubulin pri-mary antibody, washed, and incubated with goat-anti-rabbit Alexa fluor 488 secondary antibody. After wash-ing with PBS, coverslips were mounted onto glass slides using Roti-Mount FluorCare DAPI (Roth) and imaged with a Leica SP8 confocal microscope with a 63x glycerol objective (DAPI: 405 nm, tubulin: 488 nm). Z-stacks (step size: 0.33 µm) were projected using Fiji and gamma values were adjusted for visual-ization.

References (1) Wu, Y. I.; Frey, D.; Lungu, O. I.; Jaehrig, A.; Schlichting,

I.; Kuhlman, B.; Hahn, K. M. A Genetically Encoded Photoactivatable Rac Controls the Motility of Living Cells. Nature 2009, 461 (7260), 104–108. https://doi.org/10.1038/nature08241.

(2) Tas, R. P.; Chen, C.-Y.; Katrukha, E. A.; Vleugel, M.; Kok, M.; Dogterom, M.; Akhmanova, A.; Kapitein, L. C. Guided by Light: Optical Control of Microtubule Gliding Assays. Nano Lett. 2018, 18 (12), 7524–7528. https://doi.org/10.1021/acs.nanolett.8b03011.

(3) Adikes, R. C.; Hallett, R. A.; Saway, B. F.; Kuhlman, B.; Slep, K. C. Control of Microtubule Dynamics Using an Optogenetic Microtubule plus End–F-Actin Cross-Linker. J Cell Biol 2018, 217 (2), 779. https://doi.org/10.1083/jcb.201705190.

(4) Janke, C. The Tubulin Code: Molecular Components, Readout Mechanisms, and Functions. The Journal of Cell Biology 2014, 206 (4), 461. https://doi.org/10.1083/jcb.201406055.

(5) Dumontet, C.; Jordan, M. A. Microtubule-Binding Agents: A Dynamic Field of Cancer Therapeutics. Nat Rev Drug Discov 2010, 9 (10), 790–803. https://doi.org/10.1038/nrd3253.

(6) Peterson, J. R.; Mitchison, T. J. Small Molecules, Big Impact: A History of Chemical Inhibitors and the Cyto-skeleton. Chem. Biol. 2002, 9 (12), 1275–1285. https://doi.org/10.1016/S1074-5521(02)00284-3.

(7) Kingston, D. G. I. Taxol, a Molecule for All Seasons. Chem. Commun. 2001, No. 10, 867–880. https://doi.org/10.1039/B100070P.

(8) Castle, B. T.; Odde, D. J. Optical Control of Microtubule Dynamics in Time and Space. Cell 2015, 162 (2), 243–245. https://doi.org/10.1016/j.cell.2015.06.064.

(9) Borowiak, M.; Nahaboo, W.; Reynders, M.; Nekolla, K.; Jalinot, P.; Hasserodt, J.; Rehberg, M.; Delattre, M.; Zah-ler, S.; Vollmar, A.; et al. Photoswitchable Inhibitors of Microtubule Dynamics Optically Control Mitosis and Cell Death. Cell 2015, 162 (2), 403–411. https://doi.org/10.1016/j.cell.2015.06.049.

(10) Rastogi, S. K.; Zhao, Z.; Barrett, S. L.; Shelton, S. D.; Zafferani, M.; Anderson, H. E.; Blumenthal, M. O.; Jones, L. R.; Wang, L.; Li, X.; et al. Photoresponsive Azo-Combretastatin A-4 Analogues. Eur. J. Med. Chem. 2018, 143 (Supplement C), 1–7. https://doi.org/10.1016/j.ejmech.2017.11.012.

(11) Hüll, K.; Morstein, J.; Trauner, D. In Vivo Photopharma-cology. Chemical Reviews 2018, 118 (21), 10710–10747. https://doi.org/10.1021/acs.chemrev.8b00037.

(12) Zenker, J.; White, M. D.; Gasnier, M.; Alvarez, Y. D.; Lim, H. Y. G.; Bissiere, S.; Biro, M.; Plachta, N. Expand-ing Actin Rings Zipper the Mouse Embryo for Blastocyst Formation. Cell 2018, 173 (3), 776–791. https://doi.org/10.1016/j.cell.2018.02.035.

(13) Zenker, J.; White, M. D.; Templin, R. M.; Parton, R. G.; Thorn-Seshold, O.; Bissiere, S.; Plachta, N. A Microtu-bule-Organizing Center Directing Intracellular Transport in the Early Mouse Embryo. Science 2017, 357 (6354), 925–928. https://doi.org/10.1126/science.aam9335.

(14) Singh, A.; Saha, T.; Begemann, I.; Ricker, A.; Nüsse, H.; Thorn-Seshold, O.; Klingauf, J.; Galic, M.; Matis, M. Po-larized Microtubule Dynamics Directs Cell Mechanics and Coordinates Forces during Epithelial Morphogene-sis. Nat. Cell Biol. 2018, 20 (10), 1126–1133. https://doi.org/10.1038/s41556-018-0193-1.

(15) Morstein, J.; Hill, R. Z.; Novak, A. J. E.; Feng, S.; Nor-man, D. D.; Donthamsetti, P. C.; Frank, J. A.; Harayama, T.; Williams, B. M.; Parrill, A. L.; et al. Optical Control

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 11

of Sphingosine-1-Phosphate Formation and Function. Nature Chemical Biology 2019, 15 (6), 623–631. https://doi.org/10.1038/s41589-019-0269-7.

(16) Laprell, L.; Tochitsky, I.; Kaur, K.; Manookin, M. B.; Stein, M.; Barber, D. M.; Schön, C.; Michalakis, S.; Biel, M.; Kramer, R. H.; et al. Photopharmacological Control of Bipolar Cells Restores Visual Function in Blind Mice. The Journal of Clinical Investigation 2017, 127 (7), 2598–2611. https://doi.org/10.1172/JCI92156.

(17) Thorn-Seshold, O.; Borowiak, M.; Trauner, D.; Hasse-rodt, J. Azoaryls as Reversibly Modulatable Tubulin In-hibitors. WO2015166295, 2014.

(18) Engdahl, A. J.; Torres, E. A.; Lock, S. E.; Engdahl, T. B.; Mertz, P. S.; Streu, C. N. Synthesis, Characterization, and Bioactivity of the Photoisomerizable Tubulin Polymerization Inhibitor Azo-Combretastatin A4. Org. Lett. 2015, 17 (18), 4546–4549. https://doi.org/10.1021/acs.orglett.5b02262.

(19) Sheldon, J. E.; Dcona, M. M.; Lyons, C. E.; Hackett, J. C.; Hartman, M. C. T. Photoswitchable Anticancer Ac-tivity via Trans-Cis Isomerization of a Combretastatin A-4 Analog. Org. Biomol. Chem. 2016, 14 (1), 40–49. https://doi.org/10.1039/C5OB02005K.

(20) Gao, L.; Kraus, Y.; Wranik, M.; Weinert, T.; Pritzl, S. D.; Meiring, J. C. M.; Bingham, R.; Olieric, N.; Akhmanova, A.; Lohmüller, T.; et al. Photoswitchable Microtubule In-hibitors Enabling Robust, GFP-Orthogonal Optical Con-trol over the Tubulin Cytoskeleton. bioRxiv 2019, 716233. https://doi.org/10.1101/716233.

(21) Sailer, A.; Ermer, F.; Kraus, Y.; Lutter, F.; Donau, C.; Bremerich, M.; Ahlfeld, J.; Thorn-Seshold, O. Hemithi-oindigos as Desymmetrised Molecular Switch Scaffolds: Design Control over the Isomer-Dependency of Potent Photoswitchable Antimitotic Bioactivity in Cellulo. ChemBioChem 2019, 20, 1305–1314. https://doi.org/10.1002/cbic.201800752.

(22) Alexander Sailer; Franziska Ermer; Yvonne Kraus; Re-bekkah Bingham; Ferdinand H. Lutter; Julia Ahlfeld; Oliver Thorn-Seshold. Potent Hemithioindigo-Based Antimitotics Photocontrol the Microtubule Cytoskeleton in Cellulo. ChemRxiv 2019. https://doi.org/10.26434/chemrxiv.9176747.v1.

(23) Vandestadt, C.; Vanwalleghem, G. C.; Castillo, H. A.; Li, M.; Schulze, K.; Khabooshan, M.; Don, E.; Anko, M.-L.; Scott, E. K.; Kaslin, J. Early Migration of Precursor Neu-rons Initiates Cellular and Functional Regeneration after Spinal Cord Injury in Zebrafish. bioRxiv 2019, 539940. https://doi.org/10.1101/539940.

(24) Eguchi, K.; Taoufiq, Z.; Thorn-Seshold, O.; Trauner, D.; Hasegawa, M.; Takahashi, T. Wild-Type Monomeric α-Synuclein Can Impair Vesicle Endocytosis and Synaptic Fidelity via Tubulin Polymerization at the Calyx of Held. J. Neurosci. 2017, 37 (25), 6043–6052. https://doi.org/10.1523/jneurosci.0179-17.2017.

(25) Hellal, F.; Hurtado, A.; Ruschel, J.; Flynn, K. C.; Las-kowski, C. J.; Umlauf, M.; Kapitein, L. C.; Strikis, D.; Lemmon, V.; Bixby, J.; et al. Microtubule Stabilization Reduces Scarring and Causes Axon Regeneration After Spinal Cord Injury. Science 2011, 331 (6019), 928. https://doi.org/10.1126/science.1201148.

(26) Ertürk, A.; Hellal, F.; Enes, J.; Bradke, F. Disorganized Microtubules Underlie the Formation of Retraction Bulbs and the Failure of Axonal Regeneration. J. Neuro-sci. 2007, 27 (34), 9169.

(27) Witte, H.; Neukirchen, D.; Bradke, F. Microtubule Stabi-lization Specifies Initial Neuronal Polarization. J Cell Biol 2008, 180 (3), 619. https://doi.org/10.1083/jcb.200707042.

(28) Ruschel, J.; Hellal, F.; Flynn, K. C.; Dupraz, S.; Elliott, D. A.; Tedeschi, A.; Bates, M.; Sliwinski, C.; Brook, G.; Dobrindt, K.; et al. Systemic Administration of Epothi-lone B Promotes Axon Regeneration after Spinal Cord Injury. Science 2015, 348 (6232), 347.

(29) Brill, M. S.; Kleele, T.; Ruschkies, L.; Wang, M.; Mara-hori, N. A.; Reuter, M. S.; Hausrat, T. J.; Weigand, E.; Fisher, M.; Ahles, A.; et al. Branch-Specific Microtubule Destabilization Mediates Axon Branch Loss during Neu-romuscular Synapse Elimination. Neuron 2016, 92 (4), 845–856. https://doi.org/10.1016/j.neuron.2016.09.049.

(30) Sengottuvel, V.; Leibinger, M.; Pfreimer, M.; Andre-adaki, A.; Fischer, D. Taxol Facilitates Axon Regenera-tion in the Mature CNS. J. Neurosci. 2011, 31 (7), 2688–2699. https://doi.org/10.1523/jneurosci.4885-10.2011.

(31) Zhang, Y.-M.; Zhang, N.-Y.; Xiao, K.; Yu, Q.; Liu, Y. Photo-Controlled Reversible Microtubule Assembly Me-diated by Paclitaxel-Modified Cyclodextrin. Angewandte Chemie International Edition 2018, 57 (28), 8649–8653. https://doi.org/10.1002/anie.201804620.

(32) Stricker, L.; Fritz, E.-C.; Peterlechner, M.; Doltsinis, N. L.; Ravoo, B. J. Arylazopyrazoles as Light-Responsive Molecular Switches in Cyclodextrin-Based Supramolec-ular Systems. J. Am. Chem. Soc. 2016, 138 (13), 4547–4554. https://doi.org/10.1021/jacs.6b00484.

(33) Alushin, G. M.; Lander, G. C.; Kellogg, E. H.; Zhang, R.; Baker, D.; Nogales, E. High-Resolution Microtubule Structures Reveal the Structural Transitions in Αβ-Tubu-lin upon GTP Hydrolysis. Cell 2014, 157 (5), 1117–1129. https://doi.org/10.1016/j.cell.2014.03.053.

(34) Kingston, D. G. I. The Chemistry of Taxol. Pharmacol-ogy & Therapeutics 1991, 52 (1), 1–34. https://doi.org/10.1016/0163-7258(91)90085-Z.

(35) Lukinavicius, G.; Reymond, L.; D’Este, E.; Masharina, A.; Gottfert, F.; Ta, H.; Guther, A.; Fournier, M.; Rizzo, S.; Waldmann, H.; et al. Fluorogenic Probes for Live-Cell Imaging of the Cytoskeleton. Nat Meth 2014, 11 (7), 731–733. https://doi.org/10.1038/nmeth.2972.

(36) Löwe, J.; Li, H.; Downing, K. H.; Nogales, E. Refined Structure of Αβ-Tubulin at 3.5 Å Resolution11Edited by I. A. Wilson. Journal of Molecular Biology 2001, 313 (5), 1045–1057. https://doi.org/10.1006/jmbi.2001.5077.

(37) Mitchison, T. J. The Proliferation Rate Paradox in Anti-mitotic Chemotherapy. Molecular Biology of the Cell 2012, 23 (1), 1–6. https://doi.org/10.1091/mbc.E10-04-0335.

(38) Tron, G. C.; Pirali, T.; Sorba, G.; Pagliai, F.; Busacca, S.; Genazzani, A. A. Medicinal Chemistry of Com-bretastatin A4: Present and Future Directions. J. Med. Chem. 2006, 49 (11), 3033–3044. https://doi.org/10.1021/jm0512903.

(39) Weston, C. E.; Richardson, R. D.; Haycock, P. R.; White, A. J. P.; Fuchter, M. J. Arylazopyrazoles: Azohete-roarene Photoswitches Offering Quantitative Isomeriza-tion and Long Thermal Half-Lives. J. Am. Chem. Soc. 2014, 136 (34), 11878–11881. https://doi.org/10.1021/ja505444d.

(40) Lin, C. M.; Singh, S. B.; Chu, P. S.; Dempcy, R. O.; Schmidt, J. M.; Pettit, G. R.; Hamel, E. Interactions of Tubulin with Potent Natural and Synthetic Analogs of the Antimitotic Agent Combretastatin: A Structure-Activity Study. Molecular Pharmacology 1988, 34 (2), 200–208.

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

Müller-Deku et al. 2019, AzTax | 12

Acknowledgements This research was supported by funds from the

German Research Foundation (DFG: SFB1032 Nanoagents for Spatiotemporal Control project B09 to D.T. and O.T.-S.; SFB TRR 152 project P24 number 239283807, Emmy Noether grant TH2231/1-1, and SPP 1926 project number 426018126 to O.T.-S.) and the National Institutes of Health (Grant R01GM126228 to D.T.). We thank F. Ermer and M. Borowiak (LMU) for initial MTT viability assays, P.A.S (LMU) for initial synthesis, H. Harz for micros-copy access (LMU microscopy platform CALM).

Author Contributions A.M.-D. performed synthesis, photocharacteri-

sation, and coordinated data assembly. K.L. performed cell biology. Y.K. performed initial cell biology. C.H. performed flow cytometry. R.B. performed in vitro tu-bulin polymerisation assays. J.A. supervised cell biol-ogy and coordinated data assembly. D.T. designed the concept and supervised synthesis. O.T.-S. designed the study, performed and supervised synthesis, supervised all other experiments, coordinated data assembly and wrote the manuscript with input from all authors.

Additional Information Supporting Information (PDF) accompanies this

paper: (i) synthetic protocols; (ii) photocharacterisa-tion; (iii) biochemistry; (iv) cell biology; (v) NMR spectra.

Competing Interests The authors declare no competing interests. Dedication We dedicate this paper to Tim Mitchison,

whose passion for microtubule biology, support of the field, and insistence on the role of small molecule rea-

gents within it, gave much inspiration to this work.

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

ON

360nm

OFFdark

OH O OH

OOAc

H

O

OHPh

NH

OHO

OBz

NN

R2

R1

R2

O

R3

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

OR’ OMe

OH

OOAc

H

O

OHPh

NHR

OHO

OBzEDCIHOBt DIPEA

OH OMe

OH

OOAc

H

O

OHPh

NH

OHO

OBz

NN

R2

R1

R2

O

CO2H

Azo R = Boc: docetaxel (R’ = H)R = Bz: paclitaxel (a; R’ = Ac)

1. TFAdocetaxel:

b

a

c

R3

AzTaxseries

o/m/p

2.

compound attach R1 R2 R3

AzTax4H para H H HAzTax3H meta H H HAzTax4MP para OMe H HAzTax3MP meta OMe H HAzTax2MP ortho OMe H HAzTax3TM meta OMe OMe HAzTax3MTM meta OMe OMe OMeAzTax4DMA para NMe2 H HAzTax3DMA meta NMe2 H HAzTax3DEA meta N(C2H4OH)2 H H

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

AzTax3MPPSS equilibria

AzTax3MPcell viability

tubulinpolymerisation

-7 -6 -50

50

100

log10([AzTax3MP]) (M)

cell

viab

ility

(%) dark360 nm

a b

300 400 5000.00

0.25

0.50

0.75

wavelength (nm)

Abs

dark530 nm515 nm490 nm435 nm410 nm375 nm

dEC50 ratiocompound 360nm dark dark / litAzTax4H 0.048 0.082 1.7AzTax3H 0.64 0.89 1.4AzTax4MP 2.1 2.0 1.0AzTax3MP 0.24 1.4 5.8AzTax2MP 1.8 6.8 3.7AzTax3TM 0.53 1.6 3.0AzTax3MTM 0.65 1.2 1.9AzTax4DMA 0.13 0.15 1.1AzTax3DMA 0.20 0.096 0.5AzTax3DEA 4.3 4.7 1.1

EC50 (µM)c

0 1 2 3

0.1

0.5

time (min)

Abs

340

cosolvent

docetaxel

E-AzTax3MPlit AzTax3MP

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint

S G2/MG1sub-G1

cosolvent1.5 µM dark1.5 µM 360 nm0.1 µM docetaxel

0

50

100

%PI-pos

itivece

lls

AzTax3MP cell cycleAzTax3MP cell cycle

sub-G1

S

G1

G2/M

0 0.1 0.75 1.5 3.0d λ d λ d λ d λλ

C [µM]d -7 -6.25 -5.75

0

20

40

60

80

%ce

llsinG2/M

0.1 µM docetaxel

cosolvent control

360 nm

dark

log10([AzTax3MP]) (M)

G2/M arrestb c d

a

pos.

360nm

dark

0.5 µM controls1 µM 1.5 µM

0.1 µM docetaxel

cosolvent

α-tubulinDAPI

was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. The copyright holder for this preprint (whichthis version posted September 23, 2019. . https://doi.org/10.1101/778993doi: bioRxiv preprint