Pyrrole Hemithioindigo Antimitotics with Near‐Quantitative Bidirectional Photoswitching that Photocontrol Cellular Microtubule Dynamics with Single‐Cell Precision

Abstract

We report the first cellular application of the emerging near‐quantitative photoswitch pyrrole hemithioindigo, by rationally designing photopharmaceutical PHTub inhibitors of the cytoskeletal protein tubulin. PHTubs allow simultaneous visible‐light imaging and photoswitching in live cells, delivering cell‐precise photomodulation of microtubule dynamics, and photocontrol over cell cycle progression and cell death. This is the first acute use of a hemithioindigo photopharmaceutical for high‐spatiotemporal‐resolution biological control in live cells. It additionally demonstrates the utility of near‐quantitative photoswitches, by enabling a dark‐active design to overcome residual background activity during cellular photopatterning. This work opens up new horizons for high‐precision microtubule research using PHTubs and shows the cellular applicability of pyrrole hemithioindigo as a valuable scaffold for photocontrol of a range of other biological targets.

Pyrrole hemithioindigos can be near‐quantitatively photoisomerised by the laser wavelengths available on confocal microscopes. The first pyrrole hemithioindigo‐based photopharmaceuticals, PHTubs, which are photoswitchable tubulin inhibitors, have been developed. PHTubs optically inhibit microtubule dynamics in live cells with single‐cell, second‐scale spatiotemporal precision, with both lit‐active and dark‐active compounds demonstrated in action.

Introduction

Optically targeted approaches to manipulate biological systems have revolutionised diverse research fields, by enabling the non‐invasive patterning of protein activity with resolution on the spatial and temporal scales inherent to cellular functions. The high‐precision studies that optical approaches allow have particular promise for systems whose spatiotemporal regulation is key to their functions: such as action potential firing in neuroscience, or cytoskeleton structure and dynamics in cell biology.[ ¹ , ² , ³ ] The major optical approaches are optogenetics, ^[4] photouncaging, ^[5] and photopharmacology. ^[6] Photopharmacology offers both distinct performance and a unique application space. Unlike photouncaging, photopharmaceuticals can be modulated reversibly, to avoid a build‐up of background bioactivity that could compromise biological readout; their photoresponse is near‐instantaneous and does not generate toxic uncaging byproducts; and their photoresponse wavelengths are often more biologically compatible and more tunable than those of caging groups.[ ⁷ , ⁸ ] Unlike optogenetics, photopharmaceuticals do not need genetic engineering, so they transition easily between biological models; and they can target proteins which do not allow functional photocontrol by fusion constructs, such as the cytoskeletal scaffolds tubulin and actin.[ ³ , ⁸ , ⁹ ] Since the spatiotemporally precise orchestration of cytoskeleton structure and dynamics is critically important to hundreds of biological processes, cytoskeleton photopharmaceuticals have emerged as a valuable goal of research in recent years.[ ¹⁰ , ¹¹ ] They hold promise for basic research into anisotropic processes, for example, in intracellular transport, mechanostasis and cell motility, as well as for applied research on, for example, anti‐invasive strategies in cancer therapy.[ ¹⁰ , ¹² , ¹³ ]

The development of photopharmaceuticals has however been held back by the limited space of photoswitch scaffolds shown to permit bioactivity photoswitching in cells, and by their drawbacks. An “ideal” photoswitch scaffold would permit near‐complete, bidirectional photoisomerisations in situ in live cells under practical conditions, to allow efficient and precise photopatterning of its photoisomers. An ideal photopharmaceutical would further decorate this scaffold to create a ligand with near‐zero bioactivity as one isomer, and potent bioactivity as the other isomer. The vast majority of cellularly applicable photopharmaceuticals rely on azobenzene scaffolds, with diarylethenes, fulgides and hemithioindigos as minor players (Figure 1 a); yet these scaffolds are far from ideal. Here we expand the space of cellular photoswitch scaffolds, tackling two performance problems that azobenzenes in particular fail to address. In doing so we provide both a useful new reagent for photocontrol of the microtubule (MT) cytoskeleton, and general advances for practical photopharmacology.

Figure 1.

Design and synthesis of PHTubs: a) Photoswitchable scaffolds azobenzene, styrylbenzothiazole, hemithioindigo, and pyrrole hemithioindigo, with typical best photostationary state isomer ratios achieved under photoswitching in the biologically compatible wavelength range, and corresponding FDR_max values. b) PHTubs designed by derivatising PHT towards the pharmacophore requirements of the colchicine binding site on tubulin. c) Structures of PHTubs synthesised and tested in this study. d) Summary synthesis of representative reagent, PHTub‐7, by piperidine‐catalysed aldol condensation of the corresponding thioindoxyl and pyrrole carboxaldehyde; PHTub‐7 is depicted using its X‐ray crystal structure (Z isomer).

The first disadvantage of current photoswitches that we focus on, is that, typically, they cannot be quantitatively bidirectionally photoisomerised with biologically compatible wavelengths, which limits their power to photopattern biological activity. Using azobenzene as the most important example, its largest typical change of photostationary state (PSS) composition is around 4–5‐fold (between ca. 15:85 trans:cis under 370 nm and ca. 80:20 trans:cis under 450 nm, though new polysubstituted or heteroaromatic azobenzenes can perform better).[ ¹ , ¹⁴ , ¹⁵ , ¹⁶ ] Thus, even if only one of an azobenzene's two isomers is biologically active, photoswitching with cell‐compatible wavelengths (>360 nm) can, in a best‐case scenario, achieve just 4–5‐fold photocontrol over the concentration patterning of the bioactive species. We define this useful quantity of the photoswitch scaffold as its maximum functional dynamic range, FDR_max [Eq. 1], where ϕ _F is the fraction of the biologically more active isomer after the most complete “forwards” photoisomerisation (PSS with the highest content of this isomer), and ϕ _B is the remaining fraction of that isomer after the most complete back‐isomerisation. Note however, that typical photopharmaceuticals have substantial residual bioactivity in the less active isomer: so their actual functional dynamic range of bioactivity (FDR) cannot approach the FDR_max, which only reflects scaffold isomerisation (Figure 1 a).

$${\displaystyle \mathrm{FDR}{}_{\max }=\frac{\phi {}_{\mathrm{F}}}{\phi {}_{\mathrm{B}}}}$$ (1)

The second disadvantage we focus on, is that the fixed‐wavelength lasers available to biologists cannot deliver these “best‐case” isomerisations, since typical confocal microscopes only have a subset of the 405, 442, 488, 514, 561, and 647 nm laser lines.

These disadvantages can combine to ruin the performance of even biologically well‐designed photopharmaceuticals. For example, the MT‐inhibiting azobenzene PST‐1 (Figure 1 b),[ ³ , ¹⁷ , ¹⁸ , ¹⁹ ] designed towards an exceptional 10 000‐fold difference of tubulin‐binding affinity between the photoisomers, ^[20] has a FDR_max of just 9‐fold with “best‐case” photoisomerisations (380/525 nm). This is still comparatively high for azobenzenes; but with biologically available laser lines, the practical FDR that can be reached is just 3‐fold (405/514 nm): a sharp loss of potential for an otherwise excellent photopharmaceutical. ^[3]

To create practical MT cytoskeleton photopharmaceuticals, we therefore focussed on photoswitch scaffolds with high FDR_max that also match to biologically accessible visible‐light lasers. Previous explorations of C=C‐based photoswitches (405 nm‐switchable styrylbenzothiazole ^[21] as well as visible‐light‐switchable hemithioindigos;[ ²² , ²³ , ²⁴ , ²⁵ , ²⁶ ] Figure 1 a) have highlighted their robustness to glutathione (GSH) ^[27] as well as their tolerance of tautomerisable polar functional groups (e.g. para‐hydroxy groups) that are important for ligand–target interactions. These features make C=C‐based photoswitches attractive for further research, particularly compared to azobenzenes, which are known to suffer GSH lability, and where tautomerisable functional groups counteract their photoisomer patterning.[ ²⁸ , ²⁹ ]

Newhouse and Zweig recently developed the C=C‐based photoswitch pyrrole hemithioindigo (PHT, Figure 1 a), which has an outstanding FDR_max of ca. 40 in organic media, due to band separation between the groundstate Z and metastable E isomers, driven by the E isomer's intramolecular pyrrole NH⋅⋅⋅O=C carbonyl hydrogen bond.[ ²⁷ , ³⁰ ] While the photostability and GSH resistance of PHT encourage biological use, their report described insolubility and poor/zero photoswitching in increasingly aqueous media. ^[27] However, we previously observed that polymethoxylated derivatives of otherwise low‐solubility scaffolds can be photoswitched repeatedly in situ in cells, to photopattern bioactivity with performance recalling that seen in apolar media. ^[31] We attributed this to cellular sequestration of the hydrophobic photoswitches primarily in water‐excluded/lipid environments that do allow photoswitching, giving comparatively slow cell‐averaged rates of spontaneous E‐to‐Z relaxation, yet with rapid exchange from the aprotic environment supplying the photopatterned isomer ratio to the cytosol. ^[31] We therefore reasoned that a sufficiently soluble pyrrole hemithioindigo photopharmaceutical might apply a cellular FDR approaching this high FDR_max, as its photostationary state equilibria (PSSs) could reflect those of apolar media.

This we determined to test experimentally, evaluating the PHT scaffold in the biological context through proof‐of‐concept applications photocontrolling the dynamics of the MT cytoskeleton. This led us to design PHTubs, PHT‐based photopharmaceuticals intended to isomer‐dependently bind tubulin at the colchicine site, for cellular use as photoswitchable antimitotics and cytoskeleton inhibitors aiming at high‐FDR and high‐spatiotemporal‐resolution photocontrol of cell cycle, cell viability, and cellular MT dynamics.

Results

Design and Synthesis

We designed PHTubs as PHT‐based close steric analogues of the MT inhibitor colchicine. The colchicine pharmacophore consists of two aryl blades: a “southern” di‐ or tri‐methoxyaryl ring, fixed close to a “northern” ring projecting a methoxy or ethyl spacefilling group (Figure 1 b). ^[32] We chose the thioindoxyl to replace the southern ring to permit flexible substitution, leaving the pyrrole ring to bear the single ethyl group. We created nine PHTubs to scan orientations of the di/trimethoxy and ethyl substituents seeking both potency and isomer‐dependency of bioactivity.

Of these, we expected that the 3′‐ethyl PHTub‐3/6/9 would not bind tubulin with high affinity in either isomer due to the misoriented projection of the ethyl group (Figure 1 b). We therefore expected PHTub‐3/6/9 would serve as useful regioisomeric controls, mismatching the SAR understood for the desired biological target. We and others ^[33] use regioisomeric controls to test for nonspecific bioactivity: both the classic effects of pan‐assay interference compounds (PAINS) that must be anticipated for hydrophobic photoswitches (e.g., promiscuous binding, aggregation on proteins, or membrane disruption); ^[34] as well as nonspecific scaffold toxicity or phototoxicity under illumination. ^[22] By excluding such activity, regioisomeric controls can support that other compounds designed to match the target SAR act isomer‐dependently and target‐specifically (see Supporting Note 1). We also synthesised an N‐methylated control PHTub‐NMe that we expected to be inactive due to a steric clash (of the N‐methyl group with the binding pocket wall), ^[21] but we were interested to study the effects of suppressing the NH⋅⋅⋅O=C hydrogen bond upon its photochemistry and, potentially, bioactivity (Figure 1 c).

These PHTubs were synthesised in four to eight steps, assembling the PHT cores by piperidine‐catalysed aldol condensation of ethylpyrrole‐2‐carbaldehydes onto methoxylated thioindoxyls (Figure 1 d, Scheme S2, see Supporting Information).

Photoswitching

The PHTubs could be photoswitched in dry or in 25 % aqueous polar media, under air, over multiple cycles without degradation (Figure 2 a, Figure S2–S5). We would expect these lipophilic PHTubs to concentrate in water‐excluded environments in cells, and so to display good cellular photoswitchability at 430–450 nm (giving ca. 90 % E) vs. 515–530 nm (giving ca. 95 % Z; Figure 2 b, Figure S8, Table S4). By low photoresponse to >550 nm we also expected that they could permit orthogonal imaging in the red fluorescent protein (RFP) channel at 561 nm.

Figure 2.

Photoswitching: a) PHTub‐7 is robustly and reversibly photoswitched from a majority‐Z to a majority‐E photostationary state (PSS) at 440 nm, and back to majority‐Z with 532 nm. b) PHTub‐7 PSS absorbance spectra (UV/Vis spectroscopy) and composition (HPLC analysis) when illuminated with biocompatible wavelengths.

Matching expectations from Newhouse and Zweig, ^[27] the band separation of PHT isomers (and thus the FDR_max) depended on the solvent's capacity to interrupt the internal H‐bond, with ca. 45 nm separation in aprotic dry MeCN (FDR_max ca. 20‐fold at 440/525 nm for all compounds) decreasing to ca. 30 nm in 75:25 DMSO:PBS (FDR_max below 10‐fold, Table S2). N‐methylated PHTub‐NMe had similar photoswitching as phenyl hemithioindigo in both apolar solvents and DMSO:PBS, highlighting the role of the NH⋅⋅⋅O=C hydrogen bond in reaching high FDR. While the 3′/4′‐ethylated PHTs had all‐Z dark‐adapted states, 5′‐ethylated PHTub‐3/6/9 seemed to have nonzero E isomer proportion in the dark‐adapted state (Figure S5 and discussion in the Supporting Information). Typical E‐to‐Z relaxation halflives were in the hours to minutes range, being fastest in hydrogen‐bond‐disrupting polar solvents (Figure S6,7). PHTubs withstood challenge by 10 mm GSH over hours, and during repeated cycles of Z–E isomerization and thermal relaxation (Figure S9). This contrasts favourably to similarly polymethoxylated Z‐azobenzenes that are destroyed by thiols. ^[21]

We therefore began exploring the applicability of PHT as an in situ‐photoswitchable scaffold for cell biology through the use of PHTubs as photoswitchable MT inhibitors.

Structure‐and‐Photoisomer‐Activity Relationship

Microtubule inhibition blocks cell proliferation, ultimately inducing cell death. ^[35] Therefore, we initially assessed the cellular bioactivity of the PHTubs by assaying their antiproliferative activity, in 44 hour assays in both HeLa cervical cancer and Jurkat T‐cell leukaemia cell lines, under regularly pulsed illuminations designed to test both E and Z PHT isomers (Table 1). ^[3]

Table 1.

IC₅₀ values of all PHTubs in HeLa and Jurkat cell lines as *majority‐Z and *majority‐E isomer mixtures.^[a]

[a] *Majority‐Z by illumination at 525–535 nm or by dark conditions; *majority‐E by illumination at 435–450 nm; 44 h incubation; illuminations pulsed at 75 ms per 15 s with <3 mW cm⁻²; three independent experiments; n.sol. indicates apparent bioactivity near the solubility limit, which we attribute to aggregation; see Supporting Information for details.

PHTub‐4/5/6 gave low‐potency effects that match expectations for weak, nonspecific toxicity of low‐solubility compounds near their aggregation threshold. We interpret this as a lack of tubulin binding, due to a steric clash of their south ring's outer methoxy group with the T7 loop of the colchicine site.[ ²⁰ , ²² ] Yet, as both di‐ and tri‐methoxyaryl rings can be tolerated in the southern binding lobe, ^[32] we expected that deleting the outer methoxy (PHTub‐1/2/3) or shifting it around the ring (PHTub‐7/8/9) could permit cytotoxicity, and both were confirmed. In these series, the position of the northern ethyl group was the major determinant of both isomers’ bioactivity, which corresponded between the sets (e.g. PHTub‐1 corresponding to PHTub‐7).

4′‐Ethyl PHTub‐2/8 achieved excellent potency: superior to all photoswitchable tubulin binders yet reported[ ³ , ²¹ , ³¹ ] and near to colchicine site binders used as drugs (PHTub‐8 IC₅₀s≤120 nm, colchicine ca. 20 nm), although their bioactivity was only slightly photoswitchable (Z‐PHTub2/8 ca. 2× as potent as the E isomers).

Shifting to a 5′‐ethyl group in PHTub‐1/7 sacrificed potency, but greatly improved the photoswitchability of bioactivity: a necessary compromise for effective photopharmaceuticals. E‐PHTub‐1 had good potency (mostly‐E‐PHTub‐1 IC₅₀ ca. 2 μm in Jurkat and HeLa), while Z‐PHTub‐1 gave only weak irreproducible bioactivity above 10 μm, which we attributed to its aggregation rather than to tubulin binding. We have observed elsewhere that the out‐of‐plane middle methoxy group of a 3,4,5‐trimethoxyphenyl motif brings significant solubility benefits compared to dimethoxyphenyl, which can suppress aggregation effects and so increase the photoswitchability of bioactivity. ^[3] We were delighted to see this in PHTub‐7, which retained similarly good potency under blue illumination (mostly‐E‐PHTub‐7 IC₅₀ ca. 3–5 μm in Jurkat and HeLa), yet gave no visible effects as the Z isomer at even 10–17‐fold higher concentrations (<10 % antiproliferative effect up to 50 μm; Figure 3 a,b).

Figure 3.

PHTub bioactivity is light‐specific and tubulin‐mediated. a,b) PHTub‐7 gave strong light‐specific antiproliferative effects with wavelength‐dependency aligning to cell‐free PSS measurements (HeLa and Jurkat cells; where for comparison, PST‐1 IC₅₀s: HeLa 1 μm (435 nm), 40 μm (dark); Jurkat 80 nm (435 nm), 20 μm (dark)). c) Immunofluorescence imaging of microtubule network architecture in HeLa cells treated with PHTub‐2/7/8 or cosolvent only, under pulsed 450 nm illumination or dark conditions. White arrows indicate mitotic arrests, hollow white arrowheads indicate multipolar spindles and other mitotic failures, filled pink arrowheads indicate remnant cytokinetic bridges (12 h incubation; α‐tubulin in green, DNA stained with DAPI in blue).

Shifting instead to a 3′‐ethyl group showed that PHTub‐1/7 are at a local optimum for the photoswitchability of bioactivity. PHTub‐3 was active (E‐IC₅₀=3 μm) but had only mediocre photoswitchability of bioactivity (3‐fold), while PHTub‐9 was similarly and not strongly active as both isomers. PHTub‐NMe, whose N‐methyl group similarly acts a spacefilling group adjacent to the double bond, was fully inactive as both isomers (Table 1).

Taken together, the ordered structure–activity relationships of the PHTubs (trimethoxy orientation and spacefill location) are consistent with those expected for colchicine site binders. ^[32] The demonstration of weakly potent light‐independent regioisomer controls, for example, PHTub‐6/9, and highly‐potent but nearly‐light‐independently active PHTub‐2/8, argues that phototoxicity or other nonspecific toxicities are not relevant for the PHTubs or the PHT scaffold. Due to its excellent photoswitchability of bioactivity, we proceeded into further biological testing with PHTub‐7 as our lead PHT photopharmaceutical. In parallel, we used PHTub‐2/8 interchangeably as low‐photoswitchability but highly bioactive controls likely to share E‐PHTub‐7′s mechanism of action.

Tubulin‐Specific Mechanism of Action in Cells

To test the tubulin‐specificity of their light‐dependent cytotoxicity, we first imaged the endogenous microtubule (MT) network architecture in cells incubated with PHTubs, to observe direct effects on their desired target protein. The nanomolar‐potent PHTub‐2/8 caused a high proportion of mitotic arrests (arrows) and multipolar spindles or other mitotic failures (hollow arrowheads) under both illumination conditions (Figure 3 c), reflecting expectations from their cytotoxicity assays. PHTub‐7 was less potent in its effects, matching the cytotoxicity results: but the 450 nm‐specific accumulation of mitotic arrests and of abnormal remnant cytokinetic bridges (filled arrowheads)—a hallmark of perturbed mitosis ^[36] —was notable, whereas this was almost not seen in dark conditions (Figure 3 c). Cell‐free assays testing inhibition of the polymerisation of purified tubulin protein supported a direct tubulin‐binding mechanism of action for the PHTubs, with PHTub‐7 and PHTub‐8 displaying comparable potency as the reference colchicine site inhibitors, nocodazole and colchicine, yet with isomer‐dependent bioactivity matching, respectively the lit‐activity and dark‐activity of the cellular assays (Figure S10). We concluded that PHTubs were supported as cellularly active inhibitors binding tubulin.

To crosscheck if the major cellular bioactivity mechanism of the PHTubs is MT inhibition, we examined their induction of light‐dependent G₂/M‐phase cell cycle arrest at their antiproliferative concentrations, which is a hallmark of microtubule inhibitors. ^[32] Using flow cytometric analysis we observed, as expected, induction of G₂/M arrest for E‐PHTub‐7 but not for Z‐PHTub‐7 (Figure 4), while PHTub‐2 gave potent but nearly light‐independent cell cycle arrest (Figure S12). This further supports that E‐PHTub‐7 potently inhibits MT function while Z‐PHTub‐7 does not, suiting it to photoswitching‐based control of MT‐dependent processes.

Figure 4.

Cell cycle photocontrol. Cells treated with PHTub‐7 under 425 nm pulsing show significant G2/M arrest and cell death (sub‐G1 phase cells), similar to that seen with reference MT inhibitor PST‐1 (12 μm) and light‐independently with nocodazole (300 nm); but PHTub‐7 has no effects under 525 nm pulsing (compare to DMSO cosolvent‐only controls).

Live Cell Photocontrol of Microtubule Dynamics by In Situ PHTub Photoswitching

We next tested the suitability of PHTubs (and by extension, the PHT scaffold) for spatially and temporally precise live cell photoswitching uses in confocal microscopy. This is a significant advance for photopharmacology, since as far as we know, no hemithioindigo derivatives have yet been demonstrated in temporally resolved live cell studies with spatially resolved in situ photoswitching. To do so, we imaged microtubule dynamics in HeLa cells transfected with fluorescently labelled end binding proteins (EBs). EBs mark the GTP cap regions of polymerising MTs. Thus, imaging EBs reveals the plus ends of polymerising MTs as dynamic comets; ^[37] and imaging during photoswitching is a spatiotemporally resolved readout for isomerisation‐dependent inhibition of MT dynamics by photoswitchable inhibitors.[ ³ , ³⁸ ]

We first confirmed that Z‐PHTub‐7 (dark state) has no significant residual effects on MT dynamics by imaging tdTomato‐EB3 at 561 nm; Z‐PHTub‐7 caused no significant change of MT dynamics compared to DMSO controls (Figure 5).

Figure 5.

PHTubs enable in situ photoswitching of MT dynamics in live cells. a–d) MT dynamics are temporally precisely suppressed when cells treated with lit‐active Z ‐PHTub‐7 (2 μm) are illuminated within the field of view (FOV) at a,b) 442 nm or at c,d) 487 nm (related to Movie S1). e,f) Cells treated with dark‐active Z ‐PHTub‐8 (0.5 μm) have suppressed MT dynamics, which can be repeatedly accelerated by illuminations at 442 nm (related to Movie S3). Photobleaching background rate fits (grey: DMSO, pink: treated) are guides to the eye to distinguish accelerations. g,h) Cell‐precise MT dynamics suppression is caused by 442 nm photoactivation of Z ‐PHTub‐7 (2 μm) within a subcellular region of interest (shown by green circle) in a target “ROI” cell (green arrowhead), while adjacent “neighbour” cells without photoactivation (pink arrowhead) do not show altered MT dynamics (related to Movie S4; EB3‐tdTomato‐transfected HeLa cells; tdTomato imaged at 561 nm; normalised EB3 “comet” counts quantify dynamic MTs; b/d/f/h: scale bars indicate 20 μm; for full caption and details see Supporting Information).

Next, we applied 442 nm laser pulses to photoisomerise PHTub‐7 in situ to the E isomer throughout the field of view. This drastically suppressed cellular MT dynamics with high temporal precision (Figure 5 a,b). MT dynamics did not recover over minutes; in light of further experiments, we believe this indicates that E‐PHTub‐7 does not rapidly diffuse out of cells because it is lipophilic (intracellular “sinks”; ^[39] see below).

442 nm is nearly an ideal photoactivation wavelength for PHTs (Table S4), but it is not a very common laser line on the confocal microscopes where MT research is performed. However, GFP‐exciting lasers (typically 487, 488, or 491 nm) are standard on all confocal microscopes. We therefore tested whether GFP lasers are suitable for cellular photoactivation of PHTub‐7. Indeed, 487 nm illumination also caused reproducible inhibition of MT dynamics (Figure 5 c,d; Movie S1), although with slower response than 442 nm. These useful results show that PHT‐based photopharmaceuticals can be photoisomerised using standard blue or cyan confocal microscope lasers, that have low potential for phototoxicity in cell research (see also Supporting Note 2). This contrasts favourably to most photocages and to most other photopharmaceuticals that require UV (355 nm laser) or near‐UV (405 nm laser) photoactivations.

To crosscheck the mechanism of action of PHTub‐7, we also imaged its potent congener PHTub‐8, which had appeared more active as its dark (Z) isomer than in its lit state. Encouragingly, we observed both that Z‐PHTub‐8 suppresses MT dynamics and that in situ 442 or 487 nm illumination caused drastic acceleration of MT polymerisation for a short while. Photoaccelerations could typically be performed at least three times before their magnitude became less impressive (Figure 5 e,f; Movie S3; see Discussion). These results match the dark‐activity of its cytotoxicity and immunofluorescence assays, so supporting that both PHTub7/8 directly and photoswitchably inhibit tubulin.

Lastly, we performed cell‐precise photoisomerisations of PHTub‐7 in single selected cells, hoping to temporally precisely block MT dynamics in those cells, without affecting their neighbour cells. In cells targeted with 442 nm laser regions (ROIs), EB3 comet counts dropped even below what had been seen with full field of view photoactivations, and did not recover over minutes, while neighbour cells remained unaffected as in non‐treated and non‐illuminated cosolvent controls (Figure 5 g,h; Movie S4). By excluding bulk photoisomerisation in the medium, this supports the idea of intracellular sinks ^[39] of slowly diffusing PHTub being responsible for cell‐persistent effects. (It also contrasts to diffusional recovery (approx. 30 s) observed for other scaffolds, such as the styrylbenzothiazole Z‐SBTub3 ^[21] ).

Thus, PHTub‐7 can inhibit MT dynamics of target cells with temporal precision on the scale of seconds and spatial precision on the single‐cell level. Taken together, these experiments also demonstrate that the PHT scaffold may more generally be uniquely suited for in situ photoswitching use in live cells with standard confocal microscope lasers, as part of both dark‐ or lit‐active photopharmaceuticals, also against other protein targets.

Discussion

Photopharmacological approaches to high‐spatiotemporal‐precision manipulation of endogenous proteins have progressed significantly, enabling biological application to membranes, ion channels, and the cytoskeleton.[ ¹ , ⁹ , ⁴⁰ ] Photopharmaceuticals for the cytoskeleton have transitioned rapidly beyond their cellular proofs of concept showing non‐invasive, reversible, cellularly resolved optical control over MT dynamics and network structure. ^[41] For example, despite the relatively poor functional dynamic range of the azobenzene PSTs (3‐fold at 405/514 nm), they have found high‐precision uses in embryonic fruit fly D. melanogaster, ^[42] worm C. elegans, ^[3] zebrafish D. rerio, ^[43] and mouse M. musculus, ^[44] where they have helped to resolve questions in development and in neuroscience.[ ⁴⁵ , ⁴⁶ ] Such uses illustrate the power of photopharmacology to enable previously inaccessible studies of spatiotemporally complex processes.

However, the scope of photopharmacology remains restricted by the photochemical and biochemical limitations of its photoswitch scaffold repertoire. ^[47] Several limitations of the major scaffold azobenzene, including the diazene's susceptibility to thiol‐mediated degradation and its restricted scope of substituents compatible with bistable photoswitch performance, are well‐known. ^[21] Non‐azobenzene photoswitches that avoid these drawbacks, such as dithienylethenes, hemithioindigos, and heterostilbenes, have only recently begun to be used in bioactive pharmacophores validated in cellular studies;[ ²¹ , ²² , ⁴⁸ ] and the development of new photoswitch scaffolds is a valuable research goal in general. ^[49] Yet, while some recently developed scaffolds have approached near‐quantitative bidirectional photoswitching in physiological media, ^[50] there is still no clear choice of scaffold that can robustly ensure this performance in cells using the light sources that are actually available in biological research.

Here we show that PHT may be a general photopharmaceutical alternative to azobenzenes, offering comparable size and chemical simplicity, but better GSH resistance, more cell‐compatible photoisomerisation wavelengths, and a better match of its optimum photoisomerisation wavelengths to those available on all confocal microscopes.

We had particularly sought to develop a photopharmaceutical with a high functional dynamic range (FDR). Towards this end, we hoped that with the PHTubs’ hydrophobicity favouring biolocalisation in water‐excluded environments, their cellular photoswitching could approach the excellent PSSs that PHT shows in organic media. ^[27] This is not yet generally recognised as a design opportunity for photopharmacology, but we believe it is an important one. Tuning photochemistry dependent on cellular/local environment has long been exploited in fluorescence microscopy ^[51] and its applications are advancing rapidly; ^[52] and environment‐dependency has been observed experimentally for photoswitches (e.g. hypsochromic shifts of optimum E‐to‐Z isomerisation wavelength in membranes; ^[53] 60 nm bathochromic shifts upon protein binding ^[54] ). The photoswitchability of bioactivity observed with PHTub‐7 in this work suggests that PHT′s favourable organic‐media photochemistry can indeed be harnessed for cellular applications. This suggests new design opportunities for photopharmaceuticals to bring high functional dynamic range photoswitching to bear in biology.

Additionally, through PHTub‐8, we have given the first demonstration of time‐resolved live cell photocontrol of a dark‐active cytoskeleton photopharmaceutical. Though its reversibility was limited, this achievement is conceptually significant: because reversible dark‐active photocontrol over biological processes using stoichiometric inhibitors is more challenging to achieve than with highly nonlinearly acting modulators (as have been used to control, for example, excitable cell firing ^[55] ). Reversible stoichiometric control requires (i) high in situ photoconversion to the less bioactive metastable isomer (i.e. high practical FDR), which is typically problematic for photoswitch scaffolds (other than PHTs); and (ii) high isomer specificity of bioactivity, which is typically problematic for their bioactive derivatives (other than some colchicinoids). Although E‐PHTub‐8 had significant residual bioactivity, its high FDR—achieved by matching the PHT scaffold's peak isomerisation efficiency to the GFP laser—was crucial for its success in achieving photoaccelerations. We also believe that enhanced dark‐active compounds based on PHTub‐8 can be useful tools for localised induction of MT polymerisation in biology, which remains a high‐value goal of reagent development.[ ⁴⁰ , ⁵⁶ ]

It was interesting that PHTub‐8 appeared somewhat reversible while PHTub‐7 did not (Figure 5 e,g). We discount three suggestions for reversibility that are usually significant for photopharmacology (diffusion into/out of cells; spontaneous E‐to‐Z relaxation; and E‐to‐Z photoreversion of PHTub8 by 561 nm imaging) due to the results of control experiments (see Supporting Note 3). We propose instead that PHTub‐8 reversibility arises since upon each 442/487 nm photoisomerisation, partial E‐PHTub‐8 unbinding from tubulin monomers “injects” tubulin into the polymerisable pool, so polymerisation rate and comet count transiently increase. This excess is then “used up” by addition to MT polymer; and over ca. 10 s the cell returns to steady state MT dynamics. Thus at each subsequent 442/487 nm photoisomerisation pulse, there is less tubulin monomer that can be injected, and rate enhancements become smaller. These considerations can have general interest for photopharmacology.

Conclusion

In this work we have applied the pyrrole hemithioindigo (PHT) photoswitch scaffold to develop a rationally designed series of photopharmaceutical reagents for in situ‐photoswitching‐based optical control over the cellular microtubule (MT) cytoskeleton: a critical biological target in urgent need of reagents enabling high‐resolution modulation assays. ^[41] The choice of the PHT scaffold was driven by its superior photoswitching performance at microscopy laser lines. We have demonstrated these PHTub reagents’ capacity to apply optical control over MT network integrity, cell division, and cell death in long‐term assays. In short‐term assays, they achieve temporally resolved, cell‐specific, optical modulation of MT dynamics in live cells. To our knowledge this is the first high‐resolution live cell application of any hemithioindigo as a photopharmaceutical pharmacophore; and its success highlights both the PHTubs’ and the PHT scaffold's promise for a range of high‐spatiotemporal‐precision biological studies. We therefore believe that PHTub‐7 in particular will be a valuable photopharmaceutical for high‐precision cytoskeleton research, with promising applications in the fields of cellular transport, mechanostasis, migration, cell division, and embryonic development.

This research also suggests that the PHT scaffold too will find a broad range of applications to biological targets that are poorly addressed by current photopharmaceuticals. These targets may be located both cytosolically, as well as in water‐excluded environments where the PHT scaffold may be even more advantageously applicable. Novel photoswitches that can be introduced into known photopharmaceutical designs to retain their on‐target bioactivity while accessing new photochemical performance are rightly recognised as high‐value targets in photopharmacology (e.g. spectrally shifted diazocines,[ ⁵⁷ , ⁵⁸ ] tetra‐ortho‐substituted azobenzenes,[ ¹⁵ , ⁵⁹ ] and azoniums ^[60] ). Therefore it is broadly significant that the similar molecular design of PHTub‐7 compared to its azobenzene analogues (PSTs) ensured similar isomer‐dependent bioactivity. Our work thus suggests that Zweig and Newhouse's pyrrole hemithioindigos ^[27] may be generally useful as a straightforward alternative to azobenzenes in photopharmacology, for powerful photopharmaceuticals with higher functional dynamic ranges under standard microscopy lasers, useful against a range of biological targets.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

Supporting Information

Supporting Information

Supporting Information

Supporting Information

A. Sailer, J. C. M. Meiring, C. Heise, L. N. Pettersson, A. Akhmanova, J. Thorn-Seshold, O. Thorn-Seshold, Angew. Chem. Int. Ed. 2021, 60, 23695.