Spironaphthoxazine switchable dyes for biological imaging† †Electronic supplementary information (ESI) available: Synthetic protocols, DFT calculations, crystal structure, and additional photo-physical and microscopy characterization. CCDC 1812758. For ESI and crystallographic data in CIF or other electronic format see DOI: 10.1039/c8sc00130h

We demonstrate that a photochromic spironaphthoxazine switch operates with excellent fatigue resistance and high conversion when irradiated at 405/561 nm in a range of media including living cells.


Table of Contents
Section S1 Synthesis of spironaphthoxazine switches and dyads S2 Section S2 General procedures S3 Section S3 Photophysical properties S3 Section S4 Solvatochromism S6 Section S5 Acquisition of the photostationary state (PSS) spectra S7 Section S6 Rates of thermal ring closure S8 Section S7 Fluorescence lifetimes S8 Section S8 Re-absorption corrected fluorescence absolute quantum yields S10 Section S9 Quantum yields of photochemical reactions S11 Section S10 FRET efficiency (E) calculations S12 Section S11 Fatigue resistance in the cuvette S13 Section S12 DFT calculations S14 Section S13 Ultrafast spectroscopy S19 Section S14 Crystal structure S23 Section S15 Cell culture and staining procedure S25 Section S16 Optical microscopy S25 Section S17 Experimental synthetic procedures S28 Section S18 References S33 Section S19 Supporting spectra S34-S49 Electronic Supplementary Material (ESI) for Chemical Science. This journal is © The Royal Society of Chemistry 2018 S2 S1. Synthesis of spironaphthoxazine switches and dyads Compound 7 was prepared in two routes: deprotection of switch SO and one-pot reaction of compound S1 with zinc-chelated nitrosonaphthol (Scheme S1). The latter approach afforded product 7 with a minor amount of impurities in 28% yield, but the purity is sufficient for the subsequent coupling reaction. Switch SO was isolated in high purity, and used in the measurements of time-resolved IR and transient absorption.
Scheme S1. Synthesis of spironaphthoxazines using a zinc chelate complex. [1] Scheme S2. Synthesis of Atto565 dyes based on a modified procedure, [2] and synthesis of Dyads 1 and 2.

S2. General procedures
All reagents were purchased from commercial sources and used as received. Solvents were procured from Honeywell, formerly Sigma Aldrich. DMF was distilled and stored over molecular sieves under argon. Column chromatography was carried out using SiO2 60 (particle size 40-63 μm, Merck, UK) as stationary phase. NMR spectra were acquired on a Bruker AVII400, AVIII400, or AVII500 instrument. 1 H NMR chemical shifts are reported in ppm and were referenced internally to residual protons in the solvent (δ = 7.26 for CDCl3; 3.31 for CD3OD). 13 C{ 1 H} NMR chemical shifts are reported in ppm and were referenced internally with respect to solvent signal (δ = 77.2 for CDCl3; 49.0 for CD3OD). Standard abbreviations indicating multiplicity were used as follows: s = singlet, d = doublet, dd = double of doublets, t = triplet, q = quartet, m = multiplet, br. = broad signal. High-resolution mass spectra (HRMS) were obtained on a Bruker μTOF instrument or a Waters GCT.
Stock solutions of all compounds were prepared at concentrations in the range 0.1-5 mM and stored at -20 °C, and thawed immediately before each experiment. Spectroscopic measurements were conducted in HPLC grade solvents. The irradiation sources were mic-LED-405 (from Prizmatix Ltd.; centered at 405 nm, FWHM = 15 nm) up to 360 mW, and mic-LED-525 (from Prizmatix Ltd; centered at 525 nm, FWHM = 60 nm) up to 69 mW. To measure the ring-opening and -closing processes, the two LEDs were joined into one output beam via a beam combiner, which was coupled to a liquid light guide (diameter = 3 mm). The liquid light guide was wired close to the cell holder and orthogonal to the beamline of the spectrometer. The irradiation was performed using Matlab programs to control light sources, intensities and interval lengths, and the number of cycles. The LED power was measured as previously described. [4] The UV-vis absorption spectra were obtained on a Perkin Elmer Lambda 20 spectrometer using quartz cuvettes from Starna (10 mm path length), and temperature was controlled by a PTP-1 peltier unit from Perkin Elmer. For characterization of photo-physical and -chemical properties, the sample concentration was adjusted to have absorbance at the excitation wavelength below 0.1 to avoid the inner filter effect. The emission spectra were obtained on a FS5 fluorescence spectrophotometer (Edinburgh Instruments). Where necessary, Matlab and OriginLab softwares were used for data treatment. Figure S1. Schematic representation of the geometrical arrangement of the sample holder in the UV-vis spectrometer, for measurements of the PSS spectrum, fatigue resistance, and photochemical ring closure.

S3. Photophysical properties
The UV-vis absorption spectrum of SO in CH2Cl2 was compared with the excitation spectrum of the same sample with detection at 560 nm. The emission spectrum was recorded with excitation at 405 nm, and fluorescence emission was measured between 420 and 800 nm.

S4. Solvatochromism of SO
In each solvent, the sample was prepared to have the maximum absorbance between 0.04 and 0.07, and the UV-vis absorption spectrum was recorded. For emission spectra, the same sample was excited at 405 nm, and fluorescence emission in the region 420 -850 nm was measured.  While the absorption spectra remain unchanged, significant bathochromic shifts of the emission wavelength are observed with increasing solvent polarity, with exceptions of solvents of hydrogen bonding capacity, e.g.

2-
propanol. This is because SO, as a polar compound, has a larger dipole moment at the excited state relative to the ground state, and the interaction with polar solvents lowers the energy of the excited state, thus resulting in a red shifted emission wavelength. In addition to solvent polarity, other interactions between compounds and solvents, e.g. hydrogen bonding, can also give rise to spectral shifts.

Estimation of the PSS composition and SO(open) spectrum
Given the quick kinetics of ring closing, standard methods (e.g. NMR or HPLC analysis) are not suitable for estimating the composition of the PSS and, therefore, it is not possible to directly calculate the spectrum of SO(open). Three approaches could be considered: (a) Assume that the PSS is 100% SO(open) and 0% SO(closed). In this case the absorption spectrum of the PSS would be the same as that of SO(open).
(b) Assume that SO(open) has no absorption at 390 nm. In this case the composition of the PSS can be easily calculated as 20% SO(closed) because the ratio of the absorption of the PSS at 390 nm divided by the initial absorption of SO(closed) at 390 nm is 0.2.
(c) If the fraction of SO(closed) in the PSS is x, where x is in the range 0-20%, then the absorption spectrum of SO(open) can be calculated by subtraction a fraction x of the absorption spectrum of SO(closed) from the absorption of the PSS ( Figure S6). Assuming that the absorption spectrum of SO(open) is a smooth curve, and that it does not have zero absorption at 390 nm, implies that the composition of the PSS is 85-95% SO(open).
The assumption of a smooth spectrum is supported by the results of the TD-DFT calculations.   In the UV-vis spectrometer, a vigorously stirred solution of SO (3.3 μM, 3 mL) at 25 °C was irradiated by 405 nm LED (2.1 W/cm 2 , 2.5 s), to establish the PSS, then the irradiation was stopped and the absorbance changes at 590 nm were monitored over time. The decay was fitted to a mono-exponential function. The results indicate that dielectric constants of solvents have important effect on the kinetic rate of thermal ring closure. Additionally, in protic polar solvents (i.e. CH3OH, EtOH, 2-propanol), the thermal relaxation is much faster even to the extent that it is within the spectrometer response time and cannot be measured, whereas non-polar solvents like toluene and cyclohexane, significantly slow down the process.   a The values at 25 °C (unless otherwise stated in parentheses) are extracted from references. [5,6] b The process was too fast to be recorded.

S7. Fluorescence lifetime measurements
The fluorescence lifetime was determined by time correlated single photon counting (TCSPC) using the FS5-TCSPC unit (Edinburgh Instruments) with a picosecond pulsed diode laser (473.4 nm ± 0.5 nm). Once a time-resolved fluorescence decay and the instrumental response function were measured, a reconvolution fit analysis was carried out using the software Fluoracle (Edinburgh Instruments) by fitting in a monoexponential function: Fit = A + Be (-t/τ) , where A is the calculated background, B is the calculated preexponential factor, and the goodness of fit is given as χ 2 .  in PBS buffer, black; instrument response function, red; mono-exponential fit, green. Bottom: residues after the fit. The residues for each of the compounds are randomly scattered around the axis y = 0 without an apparent pattern, indicating the presence of a single fluorescence decay. b) Fluorescence decay of Dyad 1 in CH3OH, exponential fit and residues after the fit. S10 Figure S10. The fluorescence decays of Dyad 2 and the mono-exponential fits. a) Top: fluorescence decay of Dyad 2 in PBS buffer, black; instrument response function, red; mono-exponential fit, green. Bottom: residues after the fit. The residues for each of the compounds are randomly scattered around the axis y = 0 without an apparent pattern, indicating the presence of a single fluorescence decay. b) Fluorescence decay of Dyad 2 in CH3OH, exponential fit and residues after the fit.

S8. Re-absorption corrected fluorescence absolute quantum yields
The absolute fluorescence quantum yields were measured using a SC-30 integrating sphere module (Edinburgh Instruments) and the re-absorption effect was corrected when possible. For 5'-Atto565 in PBS or CH3OH, Dyad 1 and 2 in CH3OH, the excitation wavelength was 520 nm, the wavelength step size 0.1 nm, and integration time 0.2 s. Three scans were repeated for both the sample solution and the solvent ( Figure  S11). The scattering region between 512 and 528 nm, and emission region between 540 and 770 nm were chosen for the calculation of the observed quantum yields.
Due to the low fluorescence quantum yields of SO, Dyad 1 and Dyad 2 in PBS, the scattering spectra were measured with a neutral density filter behind the excitation monochromator to protect the detector, and the emission spectra were obtained separately in a standard way. Specifically, for SO in CH2Cl2, 405 nm was used as the excitation light. Once the scattering region was scaled accordingly to match the overlapped region with the emission, the scattering region between 392 and 420 nm and emission between 460 and 850 nm were used to calculate the observed quantum yield, 5.1%. For Dyad 1 and Dyad 2 in PBS buffer, the excitation source was 520 nm, and similar measurements gave quantum yields of 3.6% and 4.3%, respectively.
The above procedures give the observed quantum yields, but the true fluorescence quantum yields might be higher due to re-absorption of dyes with a significant spectral overlap between absorption and emission. Practically, the emission spectra of the sample are measured in both the standard fluorescence module and the integrating sphere. Once the tails of the two spectra are normalized ( Figure S11), the difference in the areas is calculated, denoted as a = (true em./obs. em. -1), the fraction reduced by re-absorption. The reabsorption corrected fluorescence quantum yield is then calculated using the formula, . For highly fluorescent dyes, i.e. 5'-Atto565, Dyad 1 and Dyad 2 in CH3OH, the re-absorption corrected fluorescence absolute quantum yields are only slightly higher than the uncorrected values. On the other hand, it was impossible to acquire the values for the other compounds because of difficulties to get sensible a values due to their low emission. Figure S11. Representative spectra used in the calculation of the absolute fluorescence quantum yield. a) Spectra used in the calculation of the observed quantum yield using an integrating sphere, y-axis displayed in the log10 scale. PBS solution of 5'-Atto565, black; PBS blank solvent, red. The inset is the zoom-in in the scattering region, y-axis displayed in the linear scale for clarity. b) Normalized fluorescence spectra using a standard fluorescence module (black) and an integrating sphere (red), and the percentage of the area difference a, caused by re-absorption, was used to calculate the absolute quantum yield.

Quantum yield of ring opening
A cyclohexane solution of SO (10.0 μM, 2.0 mL) was placed in the sample holder of a UV-vis spectrometer and stirred at 10 °C. While it was irradiated with pulses of mic-LED-405 (28.1 mW, 100 ms pulse, 200 ms interval), a kinetic trace of absorption intensity at 560 nm was recorded and processed in Matlab to give the time-dependent absorbance change ( Figure S12).
The ring-opening quantum yield of SO was obtained using the following equation, [7] where m is the slope of the linear fit at the initial rise stage (0.329 ± 0.012 s -1 in Figure S12), V is the sample volume (2.0 mL), NAhc are the constants, P is the power intensity (28.1 ± 0.3 mW), λ is the excitation wavelength (405 nm), A is the absorbance at the excitation wavelength, εprod is the estimated molar extinction coefficient of SO(open) at 560 nm (8.9 ± 0.5 × 10 4 M -1 cm -1 ), d is the path length of the cuvette (1 cm). The molar extinction coefficient was derived based on the assumption of 90 ± 5% photoconversion from the closed to open form at the PSS. The calculation gives the quantum yield of ring opening, o = 8.0 ± 0.7%. In DCM at 10 °C a value of o = 7.7 ± 0.6% is obtained. This value is consistent with what was observed from the TR-IR measurement, i.e. 7.4%.

Quantum yield of ring closure
The photoreversion traces of ring closure monitored at 560 nm were obtained using the same conditions as those of thermal ring closure, except that the sample was given pulses of mic-LED-525 (68.5 mW, 100 ms irradiation, 400 ms interval) while recording the spectrum (Figure 4b). The monitoring of 560 nm absorbance is based on the assumption that the closed form does not absorb at this wavelength, and upon exposure to 525 nm LED, the ring closing reaction is accelerated without significant side-reactions. Thus, the disappearance of the starting material/open isomer is at the same rate as the appearance of the product/closed isomer. The trace was reconstructed in Matlab.
To calculate the ring closing quantum yield of SO, the measurements were carried out as follows: after irradiated with mic-LED-405 (150 mW, 2.5 s) to generate the open form, the sample solution was exposed to a pulse of mic-LED-525 (68.5 mW, 100 ms) to initiate photochemical reversion. After the pulse, the sample continued to ring close thermally. During the experiment, the absorbance at 560 nm was recorded as a function of time, giving a kinetic trace. The above procedure was repeated using increasing duration of 525 nm irradiation (200 ms, 400 ms etc.). Each photochemical trace was subtracted from the thermal trace, and the average of the first few values after the light scattering region in each processed photochemical trace gave the difference caused by irradiation of 525 nm of a specified period ( Figure S13). By using the equation, [7]

S10. FRET efficiency (E) calculations
The following equations were used: [8] where r is the distance between the donor and acceptor, R0 is the Förster radius (distance at which the energy transfer efficiency is 50%), κ 2 is the dipole orientation factor (assumed to be 2/3 given free rotation between the donor and acceptor), D is the fluorescence quantum yield of the donor in the absence of the acceptor (79.5%), n is the refractive index of the solvent (1.34 for PBS buffer), J(λ) is the spectral overlap integral, fD is the normalized donor emission spectrum in PBS, εA is the acceptor molar extinction coefficient, and λ is the wavelength. The distance between the switch and dye in Dyad 1 was simulated by optimizing the molecular geometry in Hyperchem with a molecular mechanics calculation in an MM + force field. With a distance of max. 15.4 Å and min. 6.8 Å, the FRET efficiency is ca. 100% for both the cases.

S11. Fatigue resistance in the cuvette
The fatigue resistance by thermal reversion is measured as follows: a cyclohexane solution of SO (10 μM, 2.0 mL) was placed in the sample holder of a UV-vis spectrometer and stirred at 25 °C. The samples were irradiated using mic-LED-405 (2.1 W/cm 2 ) for 2.5 s at the beginning of each cycle. Immediately after irradiation, a kinetic trace of absorption intensity at 590 nm was recorded until the ring closure was completed. The above procedure was repeated for > 100 cycles. The raw data were treated in Matlab to extract the maximal and minimal absorbance after 405 nm excitation for each cycle.

S12. DFT calculations
DFT calculation where performed as described before, [9] briefly: The molecular structures were optimized at the DFT B3LYP/6-31+G(d,p) level of theory using the software package Gaussian 16. [10] Solvation in CH2Cl2 was applied using the polarizable continuum model (PCM). [11] All structures are confirmed groundstate minima according to the analysis of their analytical frequencies computed at the same level, which show no imaginary frequencies. On these minima, the vertical transition energies were calculated by timedependent density functional theory (TD-DFT) at DFT cam-B3LYP/6-311+G(2d,p) level of theory with the PCM solvation in CH2Cl2. The choice of functionals and basis sets are based on the conclusions of reference. [12] The calculated geometries for switch SO in the closed and all four open isomers CTT, CTT, TTT, and TTC are presented at the end of this section together with the Cartesian coordinates. All structures are optimized at the DFT B3LYP/6-31+G(d,p) level of theory.

Isomer stabilities
The relative stabilities of the open-form isomers were calculated and are summarized in Table S3. As reported before for similar spirooxazines, [12] only the TTC and CTC forms are likely to be observed experimentally.

Calculated IR and UV-vis spectra
The calculated IR-spectra were obtained at the DFT B3LYP/6-31+G(d,p) level of theory. Similarly, the UVvis spectra were calculated as vertical transitions as described before; Figures S18 and S19.
Although the calculated spectra are bound not to be a perfect match to the experimental data, it is obvious from the comparison in Figure S18 (UV) that the different isomers of the open form will not be sufficiently different to quantify the various contributions and whether the photo-conversion proceeds through a sequential mechanism. This limitation has hampered the study of the ultra-fast dynamics of spiropyrans and spirooxazines in the past. A completely different picture exists when considering the vibrational transitions. It is apparent from Figure S21 (IR) that the isomers of the open form have noticeable differences. It follows that transient IR is the ideal technique to investigate the putative mechanism of photo-conversion of spirooxazines and spiropyrans. This is discussed in detail in the following section.    The instruments have been described in detail in references. [13,14] Briefly:

ULTRA (TR M PS-TA)
The TA measurements were performed on the ULTRA [15] setup in the Time-Resolved Multiple Probe Spectroscopy mode [13] (hereafter TR M PS) at the Central Laser Facility (STFC Rutherford Appleton Laboratories, Oxfordshire, UK). The TA experiments were driven by a 10 kHz repetition rate Ti:Sapph amplifier (Thales) as a probe source, producing 40 fs pulses at 800 nm. A fraction of the Ti:Sapph laser output was used to generate a white light continuum (WLC) in a CaF2 plate to probe the sample transient absorption. The WLC was passed through the photoexcited spot on the sample (probe spot size approx. 80 μm) and dispersed through the 0.25 m (f/4) grating spectrograph and detected using a silicon diode array (Quantum Detectors). The long-pass filter was used in front of the spectrograph to block the scattered excitation radiation. No referencing was required for the TA experiments thank to sufficiently stable white light. The excitation source for the TA experiments was the second harmonic (400 nm) of the 1 kHz titanium sapphire amplifier (Spectra Physics Spitfire XP, 100 fs pulse length), pulse energy at sample attenuated down to 0.2-1 µJ and focused down to 150 × 150 μm 2 spot). Both ULTRA amplifier and Spitfire amplifier were optically synchronized by sharing the same seed from 68 MHz Ti:Sapph oscillator. The seed beam was delayed with an optical delay line before the 1 kHz amplifier to accommodate for the 100 fs -14.7 ns time delays between pump and probe. To go beyond 14.7 ns and up to 100 μs, subsequent seed pulses are selected from the 68 MHz seed pulse train accompanied by the appropriate setting of the optical delay line. The polarization of the excitation beam at sample was set to be at 54.7° with respect to the probe in TA experiments.

LIFETime (TR M PS-TRIR)
The TRIR experiments were performed in TR M PS mode on LIFEtime setup at the Central Laser Facility STFC. [14] The LIFEtime setup is based on a dual-amplifier 100 kHz Yb:KGW laser system (Pharos & Pharos SP, Light Conversion Ltd.). The two Yb:KGW amplifiers are optically synchronized through sharing the same oscillator seed beam. The output beam of Pharos amplifier is split 50/50 and used to pump the two identical OPAs (Orpheus ONE, Light Conversion Ltd.) followed with GaSe DFG stages. The two identical OPA+DFG systems produce two independently tunable mid-IR probe beams for TRIR experiments, each covering > 200 cm -1 useable bandwidth. The Pharos SP amplifier is used to pump Orpheus HP OPA generating the excitation beam for the TRIR experiments tunable from UV to mid-IR. In the current study the excitation wavelength was set at 320, 405, and 590 nm. The repetition rate of the output beam of Pharos SP amplifier was divided down programmatically by an internal pulse picker to excitation repetition rate of 0.1-1 kHz to enable Time-Resolved Multiple Probe Spectroscopy (TR M PS) mode. Relative pump-probe timing control between the two Yb:KGW amplifiers is programmable from 100 fs to 10 ms, using a combination of oscillator roundtrip timing to achieve steps of 12 ns and translation stage optical delay of the pump similar to our previous work. [13] Here, we opted for optical delay of the OPA output but not of the amplifier seed beam. The excitation beam was focused to approx. 100 μm spot at the sample and overlapped at the sample with the two mid-IR probe beams (each probe beam is approx. 100 μm in size). The excitation pulse energy at sample was set to 0.4-1.0 μJ. The polarization of the excitation beam at the sample was at 54.7° with respect to the probe beams. After the sample, the two probe beams enter independently tunable homemade spectrographs (0.15 m, f/6) and spectra are measured on two 128-element MCT detector arrays (IR Associates). The spectrum acquired by each MCT detector is integrated and digitized by an FPAS system (Infrared Systems Development Corporation). Thanks to the excellent stability of the mid-IR probe beams, no referencing has been applied in the measurements shown here.

S20
For the TR-IR measurements, the sample solution was excited using 405 nm of 0.7 μJ (or 590 nm, 0.6 µJ) at the frequency of 100 Hz (or 1 kHz), and probed in the ca. 1200-1600 and 1400-1800 cm -1 regions at 100 KHz. For TA measurements, the sample was pumped at 400 nm (0.2 μJ, 1 KHz repetition rate), and probed in the visible region (10 KHz probe repetition rate).

Data treatment
All data sets were initially corrected for time zero and group velocity dispersion (i.e. chirp correction) prior to analysis of the data. The TR-IR data were preliminarily analyzed using the software Surface Xplorer Pro and home-made UltraView.
Single wavenumber traces were fitted using OriginLab to multiexponential functions as in the following equation: where i is either 1 or 2, x is the time (ns in the time-resolved IR data; ps in the transient absorption data), ti is the lifetime, Ai is the pre-exponential factor. Figures S22 and S23 show some examples of TRIR and TA single-wavelength data at selected time ranges where the data can be roughly fitted to exponential decays. The data were further analyzed by target analysis using a custom written Matlab script. [16] Figure S22. Fitting of TA kinetics at selected single wavelengths to exponential decays, over a selected time range of 0-100 nm. a) kinetics at 432 nm. b) kinetics at 591 nm. The lifetimes from this simple analysis are slightly different from the values obtained from target analysis. Figure S23. Fitting of TRIR kinetics at selected single wavelengths to exponential decays, over selected time ranges. a, b) kinetics at 1279 cm -1 . c, d) kinetics at 1355 cm -1 . e, f) kinetics at 1488 cm -1 . Some data fit well and the lifetimes agree with those obtained from target analysis, whereas some do not.

Additional results
While TRIR could only be carried out in deuterated CD2Cl2 because of intense absorptions from the solvents, TA allowed us to study the influence of different solvents in the ring-opening photochemistry of SO.
The photo-induced ring-opening of SO was additionally measured in acetone, DMSO, THF, and toluene and the resulting TA spectra are presented in Figure S24. Qualitatively, the TA spectra look strikingly similar in term of intermediate species and timescales. The main outlier is DMSO, which appears to hamper the formation of the photoproduct. Additionally, the shape of the excited state absorption is less defined as compared to other solvents, notable CH2Cl2 and toluene. In fact, the difference of the spectrum in DMSO is marked enough for the proposed model (see main text) to fail when attempting to simulate the data (in DMSO), suggesting a different reaction pathway. Moreover, when considering that acetone also induces the same features to a lesser extent and one can tentatively suggest that polar solvents are not ideal for this photo-conversion. From steady state measurements we know that the thermal back reaction is slower (more than 50 times slower cyclohexane to MeOH) and that the fatigue resistance is best in apolar solvents.
Target analysis of the TA spectra in acetone, THF and toluene reveals similar values for the kinetic constants (see Figure S25). Minor differences can be observed on the speciation plots, with the notable difference in accumulation of the X-intermediate for the acetone and toluene cases. Nevertheless, this differences are minor and it can be concluded that solvent has only minor influence on the photo-isomerization reaction of SO. Single crystal X-ray diffraction data were collected at 150 K using an Oxford Diffraction/Agilent SuperNovae A (Cu) X-ray source. The raw frame data were integrated and reduced using CrysAlisPro (Agilent Technologies, 2010). The structure was solved using charge flipping [17,18] with SuperFlip method. [19] It was refined by full-matrix least-squares on F 2 in CRYSTALS. [20][21][22] The crystallographic data have been deposited with the Cambridge Crystallographic Data Centre (CCDC 1812758), and copies of these data can be obtained free of charge via www. ccdc.cam.ac.uk/data_request/cif.   S25 S15. Cell culture and staining procedure Human embryonic kidney (HEK) cells were cultured in Dulbecco's modified Eagle medium (DMEM), supplemented with 10% fetal calf serum (FCS). The cells were grown to confluence at 37 °C with 5% CO2 in the presence of a poly-D-lysine-coated microscope coverslip 24 h before imaging. For staining, the dyad was added to the growth medium as DMSO stock solution to reach a final concentration of 5 μM and the cells were incubated for 15 min. The coverslip was taken out of the growth medium, rinsed with dye-free medium, and used for imaging at 37 °C.

S16. Optical microscopy
Images in Figures 7, S31 and S33 below were acquired on a Zeiss 780 confocal microscope, operated with Zen software. The excitation light sources were diode lasers (405 nm, 815 μW at 100%) and (561 nm, 920 μW at 100%). The data were imported to ImageJ (Fiji software), and the color 'magenta hot' was used to present the images.
The fatigue resistance and quenching efficiencies in live cells were analyzed and presented in Figure 8 in the manuscript. The fatigue resistance in cell imaging is defined as how many switching cycles a dyad (as an entire dyad of the switch and dye) can survive before photodegradation. This property is expressed as the normalized fluorescence of the same field of view over consecutive cycles, relative to the fluorescence readout at the first bright state. The quenching efficiency, on the other hand, describes the performance of the switch without considering the effects from possible aggregation, motion of the cells, and decomposition of the dye. This presents the extent to which the switch depletes the fluorescence in a single cycle, and is defined as QE = 1 -fluorescence of the dark state/fluorescence of the bright state of the same cycle. S27 Figure S32. Fatigue resistance of Dyad 2 in live HEK cells under a confocal microscope, using fluorescence quantification of the images in Figure S33. S28 S17. Experimental synthetic procedures 1,3,3-Trimethyl-2-methyleneindoline-5-carboxylic acid [23] was prepared according to the reported procedure.