In Vivo Photocontrol of Microtubule Dynamics and Integrity, Migration and Mitosis, by the Potent GFP-Imaging-Compatible Photoswitchable Reagents SBTubA4P and SBTub2M.

Photoswitchable reagents are powerful tools for high-precision studies in cell biology. When these reagents are globally administered yet locally photoactivated in two-dimensional (2D) cell cultures, they can exert micron- and millisecond-scale biological control. This gives them great potential for use in biologically more relevant three-dimensional (3D) models and in vivo, particularly for studying systems with inherent spatiotemporal complexity, such as the cytoskeleton. However, due to a combination of photoswitch isomerization under typical imaging conditions, metabolic liabilities, and insufficient water solubility at effective concentrations, the in vivo potential of photoswitchable reagents addressing cytosolic protein targets remains largely unrealized. Here, we optimized the potency and solubility of metabolically stable, druglike colchicinoid microtubule inhibitors based on the styrylbenzothiazole (SBT) scaffold that are nonresponsive to typical fluorescent protein imaging wavelengths and so enable multichannel imaging studies. We applied these reagents both to 3D organoids and tissue explants and to classic model organisms (zebrafish, clawed frog) in one- and two-protein imaging experiments, in which spatiotemporally localized illuminations allowed them to photocontrol microtubule dynamics, network architecture, and microtubule-dependent processes in vivo with cellular precision and second-level resolution. These nanomolar, in vivo capable photoswitchable reagents should open up new dimensions for high-precision cytoskeleton research in cargo transport, cell motility, cell division, and development. More broadly, their design can also inspire similarly capable optical reagents for a range of cytosolic protein targets, thus bringing in vivo photopharmacology one step closer to general realization.

Introduction

Biological methods and instrumentation now allow us to observe cellular structures and dynamics on the micron scale, with dynamics resolved to the scale of milliseconds, in settings from in vitro cell culture[1] to in vivo (multiorgan) animal models.[2] However, the development of tools to manipulate cellular processes with matching micrometer spatial precision and millisecond temporal precision has lagged far behind, despite its value for research.[3,4]

One biological system with particularly urgent need of such tools is the microtubule (MT) cytoskeleton.[5,6] MTs are giant tubelike noncovalent polymers of α/β-tubulin protein heterodimers that extend throughout the cytosol. They are centrally organized and can be rapidly remodeled to support hundreds of spatiotemporally regulated functions in the life of a cell. The most visible roles of MTs include force generation to maintain and change cell shape and position, and scaffolding the transport of cargos by motor proteins, including chromosomes during cell division. Tools to noninvasively manipulate MT network structure and remodeling dynamics with high spatiotemporal precision have great potential to drive biological research. However, to fulfill this potential, tools must succeed in settings from two-dimensional (2D) cell culture (e.g., cell migration and division) through to in vivo, three-dimensional (3D) systems (embryonic development and neuroscience). It is also important that tools be easily transferrable across different models and organisms.[7,8]

Optogenetics has been extremely successful in patterning ion currents and cell signaling with high spatiotemporal precision. However, until recently, only two optogenetic tools to modulate MT dynamics and network architecture were known: Wittmann’s photo-inactivatable MT-polymerizing tool π-EB1,[9] and Slep’s photoactivated plus tip recruitment system.[10] Two further optogenetic tools to photocontrol enzymatic severing of MTs have recently appeared as preprints,[11,12] highlighting the interest in methods to photocontrol microtubule dynamics, network organization, and interaction partners.

By comparison, drugs to modulate MT structure and dynamics reliably across all eukaryotes are very well studied, with their ease of being translated across models and species (no time needed for breeding and validating transgenic lines, optimizing expression systems, etc.) being a strong practical advantage over genetic approaches. Taxanes, epothilones, colchicine analogues, and vinca alkaloids are all used for nonspecific suppression of MT-dependent cellular processes, in settings from single-cell studies through to in vivo therapeutic use in humans.

Much effort has been invested to develop light-triggered analogues of these drugs, to improve the spatiotemporal precision with which their MT-inhibiting activity can be applied (toward the scale of μm and ms). Photouncaging approaches have been known for some decades;[13] while more recently, photoisomerization-based drug analogues or “photopharmaceuticals” have been developed, to elegantly avoid many of the drawbacks that limited the in vivo use of typical photouncaging methods,[14] such as toxic and/or phototoxic byproducts, requirements for <360 nm wavelengths and high light intensities, slow post-illumination fragmentation, irreversibly photosensitive stock solutions, and non-optical drug release mechanisms (e.g., enzymatic hydrolysis of cages). In the MT field, recent photoswitchable analogues of taxane,[15] epothilone,[16] and colchicinoid[5,17−22] MT inhibitors have all been applied to cellular studies. These photopharmaceuticals have enabled noninvasive, reversible optical control over MT dynamics and MT-associated downstream effects, with cell-specific spatial precision and subsecond-scale temporal precision, and have been brought to bear on research in embryology,[23] neuroscience,[24] and cytoskeleton.[5]

However, the typical photoswitch scaffolds introduce problems particularly for in vivo application, three of which this paper will focus on: (1) Metabolic stability: azobenzenes, particularly as their Z-isomers, can be reductively degraded by cellular glutathione (GSH), although the substituent-dependency of this degradation has not yet been elucidated.[25−27] This degradation reduces reagent photoswitchability as well as potency, and on timescales typical for in vivo studies (>hours) the degradation byproducts are likely to give off-target effects, particularly in metabolically active later-stage animals (tested below, and also discussed elsewhere[18]). (2) Orthogonal photocontrol and imaging: under the high intensities of focussed lasers in microscopy, azobenzenes and hemithioindigos are isomerized by excitation of the typical fluorescent proteins or labels imaged in biological studies (GFP/fluorescein at 490 nm, YFP at 514 nm; some also isomerize with RFP/rhodamine imaging at 561 nm).[18,28] Typically this leaves only the 647 nm laser channel available for orthogonal imaging during photoswitching, yet this laser typically only addresses small-molecule probes. These scaffolds’ inability to allow multichannel protein imaging during photocontrol is particularly problematic for in vivo research that extensively relies on two- or three-channel imaging with fluorescent protein fusions to resolve processes with biological specificity. (3) Practical applicability: in vivo studies must ensure sufficient delivery of photopharmaceuticals to target tissues without organic cosolvents, which are less tolerated by embryos than they are in cell culture. This requires either high-potency compounds or effective solubilizing strategies: neither of which are often seen with typically hydrophobic photoswitchable drug analogues. A separate concern is to minimize the illumination needed to photoswitch an effective amount of reagent (another problem that is more urgent for in vivo than cellular studies). Photoswitch performance at optimal but unavailable wavelengths (e.g., 260–380 nm) is irrelevant for this: tuning the scaffold’s photoresponse at the laser wavelength actually used in biological studies[18] is required, to reach both high efficiency of isomerization E(λ)[5] as well as a high photostationary state (PSS).

Novel photoswitches addressing these issues are recently gaining attention to expand the biological scope of photopharmacology.[29−31] To tackle them in the context of microtubule photocontrol, we previously introduced styrylbenzothiazoles SBTub2/3 (Figure a) as highly metabolically stable, fully GFP/YFP/RFP-orthogonal photoswitchable tubulin inhibitors (Figure b,c), with excellent photoresponse to the 405 nm laser line that is standard in confocal microscopy. SBTub3 could photocontrol MT dynamics, organization, and MT-dependent processes in live cells with reversible temporal patterning, and with cellular and even subcellular spatial precision.[18]

Figure 1

Design and synthesis. (a) The colchicinoid pharmacophore (gray shaded trimethoxyphenyl south ring and isovanillyl north ring) can be applied to various scaffolds, giving photoswitchable azobenzene-based PST and SBT-based SBTub antimitotics. Previously published SBTub2/3 lacked key interaction residues (red shaded sites). (b) Z-SBTub3 inhibits tubulin polymerization and MT-dependent processes. (c) X-ray structure of tubulin:Z-SBTub3 complex (carbons as purple spheres). The south ring is buried in β-tubulin (green); only the north ring interacts with α-tubulin at the α-T5 loop (cyan). (d) Evolved SBTub compound library used in this paper. (e) Typical synthesis of SBTubs proceeds by acetanilide sulfurization, Jacobson cyclization, and basic condensation. Phosphate prodrug SBTubA4P is further accessed by phosphoester formation and deprotection.

The metabolic stability and imaging-orthogonal photocontrol of SBTub3 were excellent features toward in vivo use, but practical limitations remained, which we determined to address in this study. Primarily, while their parent colchicinoid inhibitor combretastatin A-4 (CA4) has ca. 20 nM cellular potency, the slightly larger Z-SBTub2/3 gave only micromolar bioactivity (though note that the CA4-isosteric azobenzene analogue Z-PST-1 has only ca. 500 nM potency; Figure a). The lower the potency, the more compound and the more cosolvent are needed, which we found blocked animal model applications of both SBTub2/3 and PST-1 (see below). Thus, we prioritized a structure–activity relationship (SAR) study to improve the potency of SBTubs (Figure d). Second, the SBTubs are hydrophobic, so we prioritized fully water-soluble prodrugs for in vivo use without cosolvents. Third, while the SBT scaffold is essentially a unidirectional (E → Z) photoswitch in the biological range because its isomers’ absorption bands overlap, we were curious if tautomerizable electron-donating substituents could accelerate its spontaneous thermal Z → E relaxation, as they do for azobenzenes, toward deriving SBT-based photopharmaceuticals with appreciable rates of spontaneous “bioactivity switch-off”.[28]

Our goal was then to test whether suitably potent and soluble SBT derivatives could be used for photocontrol not only in cell culture settings where many photoswitches succeed, but over a range of in vivo multi-organ animal models, with temporally specific and cell-precise in situ photoswitching after systemic administration: settings in which no other photoswitchable reagents have so far succeeded. We now report the development of these potent, soluble, metabolically stable, GFP-orthogonal SBT-based photopharmaceuticals, and characterize the optimal ligands SBTubA4 and with tubulin:SBT X-ray crystal structures. We then showcase the unprecedented success of the fully water-soluble prodrug SBTubA4P in allowing systemic administration but local photoactivations to achieve (i) spatiotemporally specific control over MT dynamics in cell culture, (ii) long-term spatially specific control over development and migration in 3D organoid culture, (iii) short-term temporally resolved photocontrol of neural development in fly brain explant, (iv) photocontrol of embryonic development in the clawed frog, and (v) temporally reversible photocontrol of microtubule dynamics in zebrafish.

Results

Cellular Structure–Activity Optimization of SBTubs

For a colchicinoid SBTub to be cellularly effective, its Z-isomer should bind tubulin,[5,32] halting cell proliferation and inducing apoptosis;[33] while the E-isomer should have negligible tubulin binding and no other significant toxicity mechanisms over a wide concentration range where the Z-isomer is bioactive, resulting in a high lit/dark bioactivity ratio (“photoswitchability of bioactivity”).[18] Thus, we performed cellular structure–activity optimization by comparing the proliferation of cells treated with SBTubs (Table and Figure S5) under in situ pulsed illuminations with near-UV light (“lit”: mostly-Z-isomer; aim: nanomolar EC50), to their proliferation without illumination (“dark”: all-E-isomer; aim: no antiproliferative effects at concentrations of up to 100 times the lit EC50).

Table 1

Cytotoxicity EC50 Values [mostly-Z lit (hν 360–450 nm), all-E dark] and Dark/Lit EC50 Ratio; and Standard Photoactivation Working Concentration [I]WC [all in HeLa cell line; na: analysis not applicable (see Supporting Information)]

	EC50 (μM)		EC50 ratio	[I]WC
compound	hν	dark	dark/lit	(μM)
SBTub2	1.4	>25	>18	6.0
SBTub3	1.2	>25	>21	3.2
SBTubA4	0.12	3.8	31	0.21
2	0.52	>100	>190	1.6
3	0.20	15	75	0.65
4	0.67	9.3	14	1.5
5	1.4	18	13	3.2
6	23	27	na	na
7	2.6	>35	>13	na
8	1.2	>100	>83	3.0
9	1.0	48	48	2.6
10	3.0	>100	>33	10
11	1.7	>100	>58	4.3
12	0.77	64	83	1.6
13	0.42	8.8	21	1.8
14	1.0	>15	>14	2.7
SBTub2M	0.035	7.0	200	0.06
16	0.52	48	92	1.1
17	0.23	40	174	0.65
18	3.0	>15	>5	10
SBTub3P	3.4	>20	>5	7.5
SBTubA4P	0.052	0.45	8.7	0.10

We furthermore introduce a single metric, the “typically useful working concentration for cell culture” [, by which to rank photopharmaceuticals for their practical utility (Table ) and so guide optimization. [I]WC is derived by examining the dose–response curves (Figures a and S5) to find the lowest concentration at which illuminated cells experience strong inhibition while nonilluminated cells experience insignificant inhibition. We systematically define [I]WC as either the lowest concentration where the lit efficacy reaches >80% of its plateau, or the highest concentration where treatment in the dark causes <10% difference of biological effect compared to the untreated control: whichever is the lower (Figure a). Szymanski has previously proposed an alternative working concentration metric [I]opt, which emphasizes maximal illuminated efficacy, and is based on idealized dose–response curves with high Hill coefficients[4] that are rarely obtained in practice. Our empirical [I]WC instead emphasizes our experience that baseline/background inhibition of biological systems must be minimal for photopharmaceutical control to be biologically useful. For example, even if IC50 values are well separated, [I]WC values are undefined if background bioactivity is too strong for the reagent to be useful (e.g., 7, Table ). We believe that ranking by [I]WC is the most useful systematic single-value method for early-stage optimization of photopharmaceuticals’ performance and that [I]WC will find traction in the community.

Figure 2

(a) Leads SBTub2M and SBTubA4 have highly nonlinear dose–response profiles, high lit/dark ratio of bioactivity, and mid-nanomolar [I]WC values. (b, c) Photocharacterization: (b) SBTub2M is not isomerized from its all-E dark state by 488 nm illumination, but is photoswitched to majority-Z lit states by UV/violet light (78% Z at 405 nm by NMR; 1:1 phosphate-buffered saline/dimethylsulfoxide, PBS:DMSO). (c) Comparison of absorbance spectra of SBTubA4 and azobenzene PST-1 illustrates the SBT’s ideal match to 405 nm photoactivation, combining stronger 405 nm absorption, with sharper absorption cutoff above 405 nm, which makes it orthogonal to GFP (488 nm), YFP (514 nm), and RFP (561 nm) imaging.

Our first priority for in vivo use was to increase Z-SBTub potency. We first tested whether restoring the methoxy groups of the CA4 pharmacophore would increase potency despite the extra size, with SBTubA4 (1). This, like most SBTubs, was synthesized in a short sequence (Figure e) via basic aldol condensation of the derivatized 2-methylbenzothiazole with the corresponding benzaldehyde, with the 2-methylbenzothiazole being obtained via Jacobson cyclization[34] of the thioacetanilide using potassium ferricyanide (exceptions were para-aniline 7 which was obtained by Horner-Wadsworth-Emmons olefination; and 9 and 11 where 2,5-/2,7-dimethylbenzothiazoles were synthesized by Ullmann-type coupling according to Ma;[35] see the Supporting Information).

SBTubA4 was an early hit, with a ca. 10-fold more potent Z-isomer than the previous best SBTub3 and with an excellent dark/lit ratio of 30 (Figure a). We then tested if the Z-SBTubs obey similar structure–activity relationships (SAR) as known for CA4. First, we rearranged the methoxy groups in 2, reducing the Z-potency, which matched literature expectations[36] as the middle methoxy group on the south ring otherwise accepts a hydrogen bond from β-Cys239. We also flipped the SBT scaffold orientation in analogues 3–5. This was not convenient for installing the hydroxy/methoxy north ring substituent pair, so in 3 and 4 we retained only one of these groups and the compounds suffered a predictable[36] though a small loss of potency. The 7-fold potency loss when moving the methoxy group from the space-filling 6-position to the 5-position which is best occupied by a small polar substituent (compound 5) was striking and expected. We concluded that indeed Z-SBTubs obey similar SAR as CA4, which could guide further development.

At this stage, we wished to test if common strategies to accelerate thermal Z → E relaxation and redshift spectral response in azobenzenes could also be applied to SBTubs. We created close steric matches of bioactive SBTub3 but with tautomerizable ortho-catechol 6 and para-amino 7 groups. Since o-hydroxy derivatives of CA4 have entered clinical trials,[36] we were surprised that 6 gave negligible potency under lit (and dark) conditions; however, 7 retained similar light-specific bioactivity as its isostere SBTub3 (see Section ).

Continuing potency optimization, we also tested whether retaining the CA4 methoxy groups on a reduced-size photoswitch would be beneficial. We created styrylthiazole (ST) 8 which can be considered as a near-perfect isostere of CA4/PST-1 minus the small polar hydroxyl group, or as a shrunken analogue of 3. As far as we know, STs have never been used in photopharmacology before. We considered them interesting to the field as we expected them to retain the metabolic stability and GFP orthogonality of SBT, while additionally being isosteric to the better-known azobenzene: which offers attractive possibilities for adapting known azobenzenes (against other targets) into potentially biologically more applicable ST-based photopharmaceuticals. 8 was accessed by closing the thiazole, starting from a methyl cinnamoylglycinate (see the Supporting Information). Pleasingly, 8 gave strong Z-specific cellular bioactivity with ca. 100-fold photoswitchability, although its 10-fold loss of potency compared to 1 suggested that for the colchicine site, we should explore SBTubs intermediate in size between SBTub3 and 1.

Therefore, for potency optimization, we tested both adding methyl groups to SBTub3 and “walking” them around the scaffold south ring (9–13); and also deleting the north ring hydroxy/methoxy substituent pair from 1 and replacing them with smaller methyls (14–18). Since SBT Z-isomers orient their sulfur toward the inner face of the molecule,[18] we did not rely on rotations around the alkene—benzothiazole single bond to “reflect” substituents into similar positions (e.g., 9 vs 12, or 10 vs 11).

South ring derivatives 9–13 did not approach the potency of 1–5, although 10–11 still have excellent performance (∼100-fold photoswitchability of bioactivity, ∼1 μM Z-potency). However, north ring derivative SBTub2M (15) where a methyl group takes the space-filling position, had excellent Z-potency (Z-IC50 35 nM) while also having excellent photoswitchability of bioactivity (ca. 200-fold; Figure a). This makes it the most potent and the most photoswitchable of the photoswitchable antimitotics known. 16 (methyl at the small polar position) and 14 and 18 (methyl projecting to inner face, clashes with protein) were predictably weaker than SBTub2M (though 16 still has submicromolar Z activity and ca. 100-fold switching), while 17 that projects the methyl to the outer face (toward the exit tunnel) was well tolerated (0.2 μM) and retained nearly 200-fold switching. It was pleasing that this SAR also followed the SAR known for CA4, as this supports that their target and binding site are conserved.

Our second priority was to develop soluble SBTubs for in vivo use in multiorgan animal models, which do not easily tolerate even low amounts of organic cosolvents, and which may require temperatures below those of typical 2D cell cultures (e.g., <30 °C for zebrafish). We aimed to formulate water-soluble SBTub prodrugs by phenol phosphorylation—an approach that has been successful for other colchicinoid inhibitors (including clinically advanced CA4P, BNC105P, etc).[37] We had initially produced SBTub3P, the phosphate prodrug of previous lead compound SBTub3, however, its aqueous solubility was only moderate (<2 mg/mL) potentially due to aggregate formation by π-stacking of the planar compound. We synthesized SBTubA4P hoping that the out-of-plane central methoxy group would reduce π-stacking, which combined with the hydrophilicity of the three extra methoxy groups would give better solubility, as we have seen in other contexts.[6,19] Indeed, SBTA4P dissolved to at least 10 mM in water at 25 °C, which was to prove important in later assays. Both prodrugs SBTubA4P and SBTub3P had similar cellular Z/E potencies as their drug forms.

In summary, we had developed SBTubA4 and SBTub2M as mid-nanomolar antimitotics with 30–200-fold photoswitchability of bioactivity for cell culture studies (both cross-validated on A549 lung carcinoma cell line, Figure S5); and we had further developed SBTubA4P as a convenient fully water-soluble prodrug of SBTubA4 for in vivo applications. These became the focus of our further biological evaluations.

SBTub Photoswitch Performance Studies

The photoresponses of most SBTubs (1–5 and 9–20) were similar to previously reported SBTs SBTub2/3,[18] with absorption maxima and absorption cut-offs being excellently balanced for both efficient E → Z photoswitching with the common 405 nm microscopy laser, and for full orthogonality to GFP imaging (ca. 488 nm). The separated π → π* absorption maxima for E- (∼360 nm) and Z-isomers (∼330 nm) enable efficient directional E → Z photoisomerization at 360–420 nm reaching ca. 80% Z (Figure b), and extinction coefficients at 405 nm were up to twice those of similar azobenzenes, promising high-efficiency photoactivation on the confocal microscope. Importantly, E- and Z-SBT absorptions drop sharply toward zero above 410 nm (Figures c, S1, and S2) which is crucial for avoiding photoresponse to 488 nm GFP imaging under high-intensity focussed lasers, as well as with broader filtered excitation sources e.g., 490 ± 25 nm: since absorption “tails” extending far beyond band maxima can otherwise cause substantial photoswitching. (For example, 561 nm RFP imaging on the confocal microscope photoisomerizes azobenzene PST-1 despite its extinction coefficients being less than 30 cm–1 M–1.[24]) However, the SBTubs’ cutoff suggested they would indeed be GFP-orthogonal, which was later confirmed and found to be crucial for in vivo use (see below).

In biological settings (>360 nm) and on the population level, these SBTs act similarly to photoactivation probes: (i) illuminations all give similar majority-Z equilibrium photostationary states (PSSs) in the photoresponsive spectral region (360 to ca. <440 nm), and (ii) they show no significant (<2%) Z → E thermal relaxation after hours in physiological buffers at pH ∼ 7 at 25 °C (Figures S3 and S4). However, they have all of the other practical advantages of photoswitches that are relevant to most research uses[38] (no toxic/phototoxic byproducts, fast illumination response, no nonoptical drug activation mechanisms, and stocks can be quantitatively relaxed to E by warming to 60 °C, which is advantageous for stock handling over sequential assays). Their photostability to continuous illumination at >380 nm was excellent. These features are shared with previously reported SBTub2/3.[18]

We used ortho-hydroxy 6 and para-amino 7 to test if strong electron donor groups that can tautomerize to freely rotatable quinoids (C–C single bond instead of C=C), can accelerate thermal Z → E relaxation to make better-reversible SBTubs; and if these would induce spectral red-shifting. However, in a cuvette, Z-6/7 thermally relaxed only slowly (half-lives ≫ hours), and since Z-7 gave similar cytotoxicity in the long-term cellular assay as its isostere SBTub3, we concluded that its cellular relaxation was not fast on a biological timescale either. The spectra of 7 were red-shifted by nearly 60 nm compared to typical SBTs putting the E-7 absorption maximum exactly at the 405 nm laser line (Figure S1); however, as neither 6/7 brought relaxation rate benefits, we did not pursue them in this study.

Styrylthiazoles as in 8 have not yet been studied as scaffolds for photopharmaceuticals.[16] We were pleased that its isomers’ spectra were only ca. 25 nm blue-shifted compared to the SBTs (giving less efficient isomerization at 405 nm), and all other properties, such as its completeness of E → Z photoswitching, were similar to the SBTs. We believe this opens up new possibilities for GFP-orthogonal, metabolically resistant photoswitches that are isosteric to azobenzenes, and so can replace them for applications where these biologically advantageous properties are required.

For all further details, including full discussion of photoswitch performance of 6/7 and of E-SBTub toxicity, see the Supporting Information including Figures S1–S5.

SBTubs Isomer-Dependently Target Tubulin in Cells

To test their cellular mechanism of bioactivity, we first examined the SBTubs’ isomer-dependent inhibition of polymerization of purified tubulin protein in a cell-free assay. Both leads SBTubA4 and SBTub2M were noninhibiting in the all-E dark state, allowing tubulin to polymerize as in the cosolvent control (Figure a). However, in the majority-Z illuminated state, they were potent inhibitors that suppressed polymerization entirely to baseline readout levels, comparable to the archetypical colchicinoid nocodazole. Matching the antiproliferation assays, they were significantly stronger inhibitors than first-generation SBTub3 (Figure a).

Figure 3

Tubulin-specific cellular mechanism. (a) SBTubs light-dependently inhibit tubulin polymerization (turbidimetric cell-free assay; absorbance mirrors the extent of polymerization; lit indicates majority-Z-SBTub (20 μM) preisomerized to PSS at 405 nm; nocodazole control at 10 μM). (b) Cell cycle analysis of Jurkat cells treated with SBTub2M/SBTubA4P matches photoswitchable reference PST-1: with significant G2/M arrest under 405 nm pulsing (lit), but without cell cycle effects in the dark (matching cosolvent controls). (c) Immunofluorescence imaging of cells treated with SBTubA4P under pulsed 405 nm illuminations (lit, mostly-Z) and in the dark (all-E), compared to cosolvent control (HeLa cells, 20 h incubation; α-tubulin in green, DNA (stained with 4′,6-diamidino-2-phenylindole, DAPI) in blue). (d) Close-up views at the colchicine-binding site of X-ray co-crystal structures (PDB 7Z01, 7Z02) of Z-SBTubA4 (green) and E-SBTub2M (orange) bound to the Darpin D1:tubulin complex (dark gray α-tubulin, light gray β-tubulin in cartoon representation; Z-SBTubA4 and Z-SBTub2M in stick representation, oxygens red, nitrogen blue, and sulfur yellow). (e) Superimposition of tubulin:CA4 (white carbons; PDB 5LYJ) and TD1:Z-SBTubA4 (PDB 7Z01) shows that lead SBTubs share the same binding site as the parent natural product CA4 (see also Figure S11).

If microtubule inhibition is also the main cellular mechanism of action of Z-SBTubs, we would expect -SBTub-treated cells to show G2/M-phase cell cycle arrest due to mitotic checkpoint failure.[15] We therefore used flow cytometry-based analysis to quantify cell cycle repartition. E-SBTubs caused either no change or small changes compared to controls, whereas G2/M arrest was strongly induced by in situ lit (mostly-Z) SBTubA4P and SBTub2M (Figures b and S6).

As G2/M-phase arrest is necessary but insufficient to conclude on cellular tubulin inhibition being their major mechanism of action, we next performed confocal microscopy imaging of the MT network architecture in immunofluorescently stained cells, to directly observe tubulin-inhibiting effects. In situ illuminated SBTubA4P caused microtubules to curve and MT architecture to break down,[39] but E-SBTubs caused no disorganization at corresponding concentrations in the dark, matching cosolvent controls (Figure c). This matches the assumption that their photoswitchable cytotoxicity arises from their Z-isomers potently inhibiting MT dynamics and stability in cells. While microscopy is a qualitative method that can misrepresent population-level statistics, the quantitative cell cycle data from flow cytometry (104 cells per datapoint) as well as the qualitative match to previous SBTub work[18] give confidence to this result.

Finally, to check our design that Z-SBTubs act as colchicinoids, we crystallized the Z-SBTubA4 and Z-SBTub2M:tubulin DARPin D1 (TD1) complexes. X-ray diffraction studies showed that both Z-SBTubA4 and Z-SBTub2M indeed directly bind tubulin at the colchicine site very similarly to CA4, even with the same orientation of its substituents (Figures d,e and S11; data deposited as PDB 7Z01 and 7Z02). This explains why the SBTub SAR determined in this study (Table ) mirrors that known for CA4. It also highlights the plasticity of the binding site, which accepts such differently sized inhibitors.

Taken together, these cell-free and cellular assays support that the SBTubs act as light-dependent tubulin inhibitors in cells, with their Z-isomers binding potently at the colchicine-binding site and their E-isomers having no effects at the corresponding concentrations.

SBTub Photocontrol Enables Cell-Precise, Temporally Reversible MT Inhibition in 2D Cell Culture

Aiming later to apply SBTubs to photocontrol MT dynamics in complex models and in vivo, we now switched to using the fully water-soluble prodrug SBTubA4P. This is an important step: (i) it avoids cosolvents that can be problematic for in vivo toxicity and (ii) it prevents hydrophobic adsorption onto the matrix materials (PDMS, collagen, agarose) that are used in 2D structured surfaces, 3D cell culture/organoid models, and for embedding motile animals during long-term imaging. Avoiding adsorption is important in our experience, as hydrophobic compounds can exhibit irreproducible apparent potencies or effects in these settings, which is timewise- and ethically prohibitive for resource-intensive low-throughput animal studies.

Before performing animal work, we probed the spatiotemporal resolution that SBTubs could achieve for in situphotoswitching-based control over MT dynamics, in 2D cell culture. Using spinning disk confocal live-cell microscopy, we imaged SBTubA4P-treated cells transfected with a fluorescently labeled fusion of the MT end binding protein EB3, to directly monitor MT polymerization dynamics during photoswitching.[18] This is possible since EB3 labels the GTP-cap of MTs, so EB3-tdTomato acts as a fluorescent marker revealing the tips of polymerizing MT plus ends in cells as hundreds of dynamic “comets” that cascade through the cell at significant velocities (tdTomato excitation at 561 nm).[40] Our protocol for cell-precise photoswitching of MT dynamics, with internal controls for compound application and for photobleaching, was as follows. We imaged transfected cells before SBTubA4P application to establish untreated MT dynamics baselines, simultaneously controlling for effects of 405 nm laser pulses on single selected cells [targeted by region of interest (ROI) illumination]; then, we added E-SBTubA4P to these same cells and continued imaging the whole field of view while applying targeted pulses of the 405 nm laser to a single ROI-selected cell.

SBTubA4P enabled repeatable cycles of temporally reversible, photoswitching-induced inhibition of MT dynamics in live cells, with single-cell spatial targeting precision and second-scale onset time precision (Figure a,b and Movie S1). In SBTubA4P-treated ROI cells, within seconds upon each single-frame 405 nm pulse, polymerizing MT tips stop moving and disappear, then more slowly reappear and resume movement (best seen in Movie S1). Statistics collected over multiple independent experiments showed these inhibition spikes are highly reproducible; recovery toward uninhibited baseline has a half-life of ca. 25 s, which we attribute to the diffusion of Z-SBTub out of the ROI cell (Figure b). There were minor effects on MT dynamics in treated non-ROI neighbor cells compared to pre-treatment controls, and the 405 nm pulsing protocol alone did not cause any readout changes (Figure a).

Figure 4

Spatiotemporal control over MT dynamics in 2D-cultured HeLa cells. (a, b) MT inhibition in SBTubA4P-treated cells is initiated only upon 405 nm illumination pulses and only in ROI-targeted cells (data related to Movie S1; live-cell EB3-tdTomato comets quantify polymerizing MTs). (a) Comet count statistics are similar to cosolvent-only baseline in both ROI-pulsed-cosolvent and non-ROI-SBTubA4P conditions; ROI-SBTubA4P statistics show inhibition spikes. (b) Stills from Movie S1 at the times indicated in (a), initially during the untreated timecourse, then during the SBTubA4P-treated timecourse on the same cells. Purple arrowhead indicates the ROI cell; purple dotted circle indicates where the 405 nm ROI is applied at times 26, 88, and 148 s; and white arrowhead indicates the non-ROI cell quantified as the internal control (scale bar 15 μm). (c) EB3 comet counts of cells imaged at 561 nm only (dark, gray), with 47 frames at 487 nm applied to full field of view during the time span indicated with dashed lines (“487”, cyan) and SBTubA4P (6 μM), or with single-frame 405 nm pulses, SBTubA4P (0.6 μM), applied to full field of view at times indicated with dashed lines (“405”, violet) (n = 3 cells). Temporally precise onset and full-field diffusional reversibility are shown (data related to Movies S2 and S3). [(a, c) Mean ± standard error of the mean (SEM) EB3 comet counts as normalized to the means of the first five time points; 405 nm ROIs applied at indicated times; for further details, see the Supporting Information].

We also modified this protocol to apply single-frame 405 nm pulses to the whole field of view instead (Figure c and Movie S2). As expected, this confirmed the temporal precision of onset and the temporal reversibility of the spiking seen with the single-cell-resolved studies; though the video data are easier to interpret as they are even more visually impressive (Movie S2). Finally, we performed full-field-of-view imaging while applying a train of 47 frames of 487 nm pulses, to test whether SBTubA4P can be used orthogonally to GFP imaging wavelengths. Indeed, even at this high concentration (6 μM, 40× [I]WC), there was no induction of MT inhibition, so we concluded that SBTubA4P is indeed GFP-orthogonal, matching our design (Movie S3 and Figure c).

From Cell Culture to MT Photocontrol in 3D Models, Tissue Explants, and Animals

By now we had optimized the potency of metabolically stable, GFP-orthogonal photoswitchable SBTub2M and fully water-soluble analogue prodrug SBTubA4P, clarified the SBTubs’ tubulin-specific cellular mechanism of action, and shown high-spatiotemporal-resolution photocontrol of MT dynamics in 2D cell culture. We were now primed to tackle the central photopharmacology research challenge, which has so far frustrated essentially all prior approaches: in vivo translation using systemic administration but local photocontrol, that clearly and usefully retains a defined cellular mechanism of action. We set out to test if SBTubs’ performance features would allow this operation across a range of complex models from 3D culture, to 3D tissue explant, to two in vivo animal models: with spatiotemporally localized illuminations photocontrolling the full sequence of their bioactivity (from suppressing MT polymerization to altering/depolymerizing MT network structures and to reducing/stopping microtubule-dependent downstream processes), where appropriate with second-level resolution, but all with cellular precision.

SBTub Photocontrol in 3D Organoids Enables Spatially Targeted Blockade of Migration and Mitosis in the Long Term

We first tested SBTubA4P in 3D human mammary gland organoids grown from isolated patient tissue embedded in collagen gels. These resemble miniaturized and simplified organs with realistic micro-anatomy, and feature collective motility/invasion behavior directing cells to migrate and proliferate to form ordered, branched structures.[41,42] Controlling organoid morphology is a sought-after goal, which has been mostly interpreted as requiring spatiotemporal control of gene expression, for which optogenetic approaches have been suggested.[43] Yet, the spatiotemporally localized application of photochemical compounds offers an alternative, in which the possibility of instantaneous cellular response to stimulus is highly attractive for temporally precise control. Based on the good performance of SBTubA4P in 2D cell culture, we assessed whether SBTubA4P can be controlled similarly precisely in a 3D organoid model, aiming to manipulate organoid development by locally interfering with the invasion of individual branches during the elongation phase.[42]

Though the actin cytoskeleton is the major driver of cell migration, MTs are integral to directional migration and leading-edge stabilization[24,44,45] (as well as to proliferation): so we expected the single-cell motility as well as cell division rates could be locally affected. Therefore, we wished to test if repeated localized photoactivations of SBTubA4P (every 7 min) could be used over long timescales (>24 h) to inhibit outgrowth of light-targeted organoid branches, while leaving other branches of the same organoids to develop: so shaping and modulating cell migration and invasion with spatiotemporal control. This aim brings conceptual challenges for photopharmaceuticals: since all cells are exposed to the same drug concentration, and over long timescales the cumulative impacts of scattered photoactivation light, of the diffused isomerized compound, and of imaging light itself, may build a spatially nonspecific background pattern of bioactivity. Organoid morphology does not tolerate >0.1% organic cosolvent either, making full solubility important.

We first determined a suitable working concentration for SBTubA4P without spatially resolved activation, by monitoring organoid areas for morphological disruption under lit/dark conditions (Figure S7a). We determined an [I]WC of 200 nM for preventing organoid growth over 24 h with UV-lit SBTubA4P, whereas organoids treated at 400 nM but kept in the dark grew healthily with no antiproliferative/branch-retracting effects (Figures a and S7a,b).

Figure 5

Spatiotemporal control over MT architecture, migration, and mitosis in 3D culture and tissue explant. (a) 3D human mammary gland organoids embedded in collagen gels only have inhibited branch outgrowth when treated with both SBTubA4P and UV pulses. (b) Local applications of UV light to ROItarg regions of SBTubA4P-treated organoids (blue box, one ca. 450 ms pulse per 7 min per z-stack) stops branch proliferation and outgrowth (red outline), while branches in untargeted ROIctrl regions develop dramatically (start: solid green line, final: dotted green line) (related to Movie S4). (c) Radial progress of branch tip fronts (directed and collective behavior) in ROItarg and ROIctrl regions. (d) Still image timecourse, zoomed on a branch tip in the ROIctrl (blue box) region, showing cell proliferation (yellow arrowhead) and matrix invasion (one representative of the migrating cells is tracked over time with green arrows), while branch tip of ROItarg region has static non-proliferating cells and even slight branch retraction (red arrows) (data related to Movie S5). (e) Branch progression and proliferation are unimpeded and continuous in ROIctrl regions, while ROItarg regions are static, and branches growing into the ROItarg stop their growth (color code as in (e), data related to Movie S6). [(a–e) Cell location in organoids tracked with nuclear stain SiR-DNA imaged at 647 nm]. (f) Whole-field-of-view 405 nm photoactivation of SBTub2M-treated intact 3D brain explants of larval Drosophila melanogaster (bottom row) causes neuroblast centrosomes (red arrows) to rapidly shrink in size and signal intensity (45 s and 3 min) and prevents the cell from progressing through division (13 min). Some MT signal accumulates at mid-cell at later time points (purple arrow) (data related to Movie S8). In DMSO-only controls (top row), centrosome integrity (white arrows, 45 s and 3 min) and progression through the cell cycle (13 min) are unaffected, indicated by myosin accumulation at the cleavage furrow (cyan arrows) (data related to Movie S9). [MTs in white (Jupiter::mCherry imaged at 561 nm), myosin in green (Squash::GFP imaged at 488 nm)]. (g) Relative mCherry fluorescence intensity of centrosomal ROIs in SBTub2M-treated prophase neuroblasts (red) after activation at 405 nm drops notably during the approximately 45 s activation period (blue box) compared to the DMSO control prophase neuroblasts (black). Signal intensities are shown as the proportion of the per-cell maximum preactivation signal intensity (shading indicates ±1 standard deviation, 1–2 centrosomes quantified from a total of five neuroblasts from three different animals). For details, see the Supporting Information.

Then, we applied localized UV ROI illuminations to selected organoid branches (ROItarg) every 7 min, comparing to non-UV-illuminated internal control branches (ROIctrl). This allowed us to noninvasively block cell migration/invasion and proliferation with striking spatial resolution and long-term persistence. With SBTubA4P, the development of ROItarg branches was totally blocked over >1 day (Figure b–d, Movies S4 and S5), while no-compound controls showed no photoinhibition of branch development (Movie S7 and Figure S7c). Control branches displayed high motility and matrix invasion (green outline, Figure b) and their cells freely proliferated (yellow arrowhead, Figure d), resulting in considerable branch development. The out-directed motion of the collective branch fronts was continuous in ROIctrl regions (ca. 4.5 μm/h) but was almost completely stopped in ROItarg (ca. 0.2 μm/h; Figure c). Overall development of branches, only entering ROIctrl areas, was however clearly visible without any statistical analysis (Figure e and Movie S6). Thus, SBTubA4P can be used in 3D matrix cell culture settings to noninvasively control cell motility, invasion, and proliferation, allowing photopatterning of branch growth and organoid development down to the spatial scale of individual cells.

SBTub Photocontrol in Intact 3D Tissue Explants Allows Temporally Precise MT Depolymerization and Mitotic Control in the Short Term

We next tested SBTub performance and tubulin-specific mechanism of action when directed against the more complex 3D environment of live intact brain lobes of early third instar larval D. melanogaster (fruit fly). Larvae are too motile for long-term imaging and the larval cuticle is largely impermeable, so neurodevelopment studies explant the whole brain. As the explant tolerates cosolvent, we took the opportunity to use SBTub2M in these assays to test the broader applicability of the SBTub design. Both the whole-organ and explant aspects bring significant challenges. (1) Compounds must permeate through two glia cell layers to reach mitotically active neural stem cells (neuroblasts). This forces the use of high bath concentrations and potent compounds: however, since surface and surrounding cells are exposed to far higher concentrations than central cells, only potent compounds with extremely high photoswitchability of bioactivity (“FDR”; discussed in ref (19)) can be used: otherwise, outer cells die, and morphology and physiology are lost. Indeed, we could use high concentrations of high-FDR SBTub2M (30 μM) in brain explants without noticeable toxicity. (2) Using multiple fluorescent protein labels for multiplexed imaging is a typical requirement to achieve useful readouts in biology, but can block chemical photoswitch applications. The most common long-wavelength fluorescent proteins for animal work are excited at 561 nm, which forces the use of GFP (488 nm) or YFP (514 nm) fusions as the next-longest-wavelength markers. For example, to image both MTs and cellular structural elements, we used animals expressing the microtubule-binding protein mCherry::jupiter,[46,47] and cortical structure marker sqh::GFP (spaghetti squash, the regulatory light chain of the non-muscle type 2 myosin, fused to eGFP[48]). With hemithioindigo or azobenzene reagents, such two-channel FP imaging would isomerize the photoswitch throughout the sample due to photoresponse at ≤530 nm (Figure b), so destroying spatiotemporal specificity in the study zone. In contrast, the nonresponse of the GFP-orthogonal SBT to 488 nm imaging avoids any photoisomerization during typical two-channel FP imaging, allowing precise temporal control of activation in our experiments.

We transferred freshly dissected brain explants into 30 μM SBTub2M and started imaging after 30 min loading (Figure S8). We imaged in both mCherry (ex 561 nm) “red channel” and GFP (ex 488 nm) “green channel” for 15 min to establish baseline, then photoactivated SBTub2M throughout the imaging stack volume with 405 nm. Photoactivated SBTub2M depolymerized centrosome microtubules within 60 s (Figure f, Movies S8, and S9, MTs shown in white). To control for target specificity, we also used mCherry::tubulin[49] for imaging MTs, and observed similar behavior (Movies S10 and S11). As microtubules are rapidly nucleated in prophase centrosomes, we quantified the loss of centrosomal fluorescence signal after activation as a highly conservative estimate of centrosome MT depolymerization. We saw dramatic, temporally resolved signal reduction at the approx. 45 s activation period, while SBTub controls were unaffected (Figure g). We used the second FP channel to image Sqh::GFP, a marker of the cell actomyosin cortex, which plays a key role in neuroblast asymmetric division.[50,51] Normally dividing neuroblasts accumulate Sqh::GFP at the cell cleavage furrow during anaphase. Neuroblasts in which SBTub2M was photoactivated retain uniform cortical myosin, indicating mitotic arrest in the absence of mitotic spindles[50] (cyan arrows, Figure f).

Previous short-term results imaging EBs at low SBTub concentrations in 2D cell culture had illustrated only its capacity to spatiotemporally block MT polymerization (Figure ). Now, these useful results in the intact brain underlined that SBTub2M maintains its Z-isomer-specific, MT-depolymerizing mechanism of action in live tissue explant, casting SBTubs as flexible and powerful tools for cytoskeleton photomanipulation in complex 3D settings.

SBTub Photocontrol in Live Animals Enables Targeted Blockade of Embryonic Development

Encouraged by performance in 3D models, we evaluated using SBTubs for in vivo photocontrol in intact animals. First, we studied the effects of in situ photoactivations of SBTubA4P on the development of Xenopus tropicalis clawed frog embryos. During the initial 48 h of development, embryos normally transition over many division cycles from cell spheres through the blastula stage through to multi-organ tadpoles (Figure S9a). Initially, we tested the effects of SBTubA4P in the earliest stages of development, just after embryonic divisions had started, by treating 2-cell stage embryos with E-SBTubA4P (1 h loading, then medium exchange and optional in situ embryo-localized 410 nm photoactivation pulse during washout; note that this transient exposure to the SBTub parallels what could be expected for e.g., systemic i.v. administration in mammalian models—further discussion in the Supporting Information). Embryos only failed to develop morphologically over the subsequent 2 days if they had received the 410 nm photoactivation; embryos without photoactivation developed normally (Figure a,b). We also tested interfering with development at the later blastula stage (>64 cells) by a similar protocol. While subsequent morphological development was normal under lit and dark conditions (Figure S9b,c), sensorimotor responses to mechanical stimulation[52] were suppressed by lit SBTubA4P only (Figure S9e and Movie S12). Interestingly, the tubulin-inhibiting azobenzene photoswitch PST-1P (Figure S9d) light-independently suppresses sensorimotor responses. We believe the SBT’s success may reflect its greater metabolic robustness (see the Supporting Information); and at any rate, it indicated that SBTubs are suitable for in vivo use.

Figure 6

Photoinhibition of X. tropicalis development, and in vivo photocontrol of MT dynamics in Danio rerio. (a, b) Xenopus embryos incubated with compounds for 1 h at the two-cell stage, before medium exchange optionally with 410 nm photoactivation. Embryos show irreversible development inhibition by in situ formed Z-SBTubA4 in lit conditions but had no effects in the dark or with a low concentration of SBTubA4P [(a) SBTubA4P at 5 μM; (b) development quantified by the ratio of major to minor embryo axis lengths, six embryos per condition, mean ± SEM]. (c) Development of D. rerio treated at the indicated stages for 24 h with SBTubA4P or control compounds under dark or pulsed lit (1 s/5 min) conditions. SBTubA4P (1 or 25 μM) causes morphological abnormalities only in the lit state, showing that it remains effective in vivo. (d, e) Reversible modulation of MT dynamics in 48 hpf zebrafish embryo (25 μM). (EB3-GFP in green, histone H2B in red). (data related to Movies S12–S16; see the Supporting Information).

SBTub Photocontrol in Live Animals Enables Cell-Precise Temporally Reversible Inhibition of MT Dynamics

Finally, we switched to highly spatiotemporally resolved in vivo MT-imaging studies that would test SBTubA4P’s mechanism of action and applicability in the zebrafish D. rerio, when systemically applied and maintained in the bath medium.

We first determined useful working concentrations in zebrafish, incubating 24 and 48 hpf (hours post fertilization) embryos in SBTubA4P under lit and dark conditions for 24 h. While zebrafish morphology remained unaltered in all dark SBTub treatments, 24 hpf embryos treated with lit SBTub showed major morphological changes even down to 1 μM, whereas more developed 48 hpf embryos showed similar morphological changes only at higher SBTubA4P concentrations e.g., 25 μM (Figures c and S10). Again, we compared these effects to those of azobenzene reagent PST-1P, now observing a dramatic difference: even 25 μM lit PST-1P did not interfere with development at either the 24 hpf or 48 hpf stage (Figures c and S10). This argues still more conclusively than Figure b, that the SBTub scaffold is uniquely suitable for light-controlled biological effects, compared to the previously known azobenzene scaffold. Finally, we checked the lower-potency soluble prodrug SBTub3P in the same assay; matching expectations, it caused only weak changes at 24 hpf and no visible changes at 48 hpf (Figure S10), showing the necessity of the potency optimizations we performed in this study.

Aiming to test the MT-modulating effects of SBTubs in a challenging live animal system, we therefore decided to proceed with 48 hpf zebrafish embryos, and an SBTubA4P working concentration of 25 μM. We took 48 hpf embryos coexpressing EB3-GFP and histone H2B-mRFP as a nuclear marker,[53,54] loaded them with 25 μM E-SBTubA4P for 4 h, then washed and embedded them in agarose. Imaging at 488 nm caused no suppression of EB3 comets, confirming SBTubA4P’s GFP orthogonality in vivo. However, photoactivation with the 405 nm laser at a single point caused EB3 comets to vanish rapidly in cells around the targeted region, recovering over ca. 10 min. Cells further from the targeted region were predictably less inhibited than those with direct contact to the photoactivation region. The photoactivation-recovery cycle could be repeated multiple times during imaging (Figure d,e and Movies S13–S16). Not only microtubule polymerization dynamics but also mitotic progression, could be stopped by spatiotemporally localized SBTubA4P photoactivation in vivo (Movie S16). These experiments confirm that the SBT scaffold, in general, is viable for light-triggered in vivo studies, and that SBTubA4P when applied in vivo retains its mechanism of action as a potent, light-dependent MT inhibitor with excellent spatial specificity and satisfying temporal reversibility.

Conclusions

Noninvasive optical tools to modulate microtubule dynamics, structure, and function with high precision offer unique potential in the many fields of biology impacted by the spatiotemporally resolved processes that MTs support, such as cargo transport, cell motility, cell division, development, and neuroscience. Photopharmaceutical chemical reagents are conceptually elegant optical tools in that they can be rapidly transitioned across models and settings and that they can be rationally designed for photoresponse patterns that interfere minimally with imaging while maximizing optical response to a chosen photoactivation wavelength. Locally applied photopharmaceuticals, particularly intraocularly applied reagents for action potential control in the retina, have made great progress in adult mammalian disease models.[55,56] However, reaching general uses of photopharmacology involving precise targeting of drug activity in vivo by localized in situ photoactivations following systemic administration remains an unsolved challenge for photopharmaceuticals with cytosolic target proteins. This would require combining high photoswitchability of bioactivity, high potency, metabolic robustness, aqueous solubility, and imaging orthogonality. Indeed, very few systemic in vivo applications of photopharmaceuticals have been made, and to date, no in vivo studies have combined testing a defined mechanism of drug action, with exploring the potential for cell-scale spatially resolved targeting following systemic administration.[57−59] Thus, the ability of photopharmacology to contribute useful systemically applicable reagents for organism studies has remained unclear.

In this work, we develop highly light-specific tubulin polymerization inhibitors with unprecedented applicability from 2D cell culture, through 3D culture and whole-organ explant, to systemic in vivo administration with local photoactivation. Realizing that the metabolic robustness and imaging orthogonality of a photopharmaceutical largely depend on its photoswitch scaffold, we consciously selected the recently developed SBT photoswitch and screened a SAR panel of 20 SBTubs. We optimized the potency and photoswitchability of bioactivity in lead SBTub2M, and we solubilized another lead to create SBTubA4P, which does not require organic cosolvents. SBTubs are efficiently photoactivated with the common 405 nm laser, but their sharp absorption cutoff leaves GFP, YFP, and RFP channels free for multiplexed imaging of fusion protein markers without risking compound photoactivation. This is a highly desirable feature for areas of research where photopharmacology’s optical precision can best contribute unique solutions on the cellular spatial scale (although this practical use logic does run counter to the goal of “photoswitch red-shifting” that is often cited in chemical design). We cross-validated the SBTubs’ molecular mechanism of action in cell-free, cellular, and in vivo settings. SBTubs light-dependently interfere with mitosis in cell culture, and in vivo they depolymerize mitotic spindles, ultimately blocking development. They can be optically patterned to control motility and branch development in 3D organoid cultures, and their photocontrol allows rapid response, cell-specific inhibition of microtubule dynamics in cell culture and in vivo.

These consistent results across a range of models at different scales of time, length, and biological complexity recommend the SBTubs as excellent and unique general-purpose tools for optically manipulating microtubule dynamics, microtubule structure, and microtubule-dependent processes with high spatiotemporal precision.

The proof-of-concept biological performance of the SBTubs has been very satisfying. We believe that the most valuable improvement to this system will now be to extend the temporal reversibility of inhibition (seen in 2D cell culture by diffusion to the medium with ca. 20 s half-time, Figure ) to whole-organ/whole-animal settings. In these settings, diffusion is slower to achieve reversibility (ca. 10 min), so we seek techniques for in situ bidirectional isomerization of SBTubs in our ongoing research. We believe that bidirectional photoswitching may be difficult within the biologically compatible wavelength range.[16] However, accelerating thermal relaxation to the minute scale, which is probably the most appropriate scale for 3D/in vivo applications of interest, may be feasible, and efforts are underway.

In conclusion, the SBTubs are excellent photoswitchable microtubule-depolymerizing reagents for use in cell culture, 3D culture, small explant, and early-stage animals. Their potency, flexibility, and ease of use recommend them for high-spatiotemporal-precision research across cytoskeleton biology; particularly, we feel, for cell-specific applications to motility and development, but they will also be of great interest in cargo transport, biophysics, cell polarity, neurodegeneration, and cell division. Finally, we expect that by supporting conceptual innovations in photoswitch scaffold chemistry and rational photopharmaceutical design, and particularly by starting to unlock the applications promise of photopharmacology for globally administered, locally targeted in vivo use, this SBTub research represents a promising advance for high-performance photopharmacology against other protein targets in general, beyond their immediate impact on microtubule biology.