Design & Synthesis

The design of our photoactivatable probe is based on several concepts. The first one relies on the fact that the leuco-Rhodamine (LRhod), the reduced form of a rhodamine is no longer fluorescent (OFF state) due to the disruption of the conjugation between the two phenyl rings of the xanthene moiety (Figure 1a). However, in oxidizing conditions, this LRhod can be reversibly oxidized into fluorescent rhodamine (ON state), Rhod. The second concept relies on the fact that, upon excitation, a fluorophore can generate singlet oxygen (¹O₂) through its excited triplet state. Consequently, we hypothesized that LRhod could get oxidized by a neighboring rhodamine, which acts as a photosensitizer (Figure 1a). To favor this mechanism, which relies on diffusion, we targeted LRhod to the plasma membrane. In the initial state, a small number of Rhod-PM molecules are already present on the cell surface (Figure 1b) due to ambient triplet oxygen, which is reactive enough to slowly oxidize leucorhodamines.^[43] However, in the plasma membrane (PM) the probes freely diffuse, which favors the interaction between the photoactivatable probe LRhod-PM and the singlet oxygen generated by its oxidized cognate photosensitizer Rhod-PM. This interaction results in the photoactivation of LRhod-PM and since the ¹O₂ generated can be responsible for the photobleaching,^[44] it can simultaneously lead to photobleaching of the emissive form (Figure 1c).

Overall, we thus expected that an efficient blinking system could be obtained in live SMLM imaging as the LRhod-PM probe alternatively turns ON and OFF, respectively, by photoactivation and photobleaching through an STPC mechanism.

To obtain the PM probes, we first synthesized a functionalizable tetramethyl rhodamine Rhod-COOH possessing a carboxylate group (Scheme 1). The latter was conjugated to dipropargylamine to obtain Rhod-DA, which was either reduced by sodium borohydride to provide LRhod-DA or clicked to clickable amphiphilic zwitterion (CAZ) to get Rhod-PM, an approach giving rise to efficient fluorescent plasma membrane probes.^{[45, 29, 46-49]} As a final step, Rhod-PM was reduced with sodium borohydride into the non-fluorescent plasma membrane probe leuco-rhodamine LRhod-PM.

Photophysical Properties

First, the photophysical properties of Rhod-PM and its precursors Rhod-DA and LRhod-DA were evaluated by spectroscopy and are reported in Figure 2. As expected, LRhod-DA did not absorb nor emit in the visible range (Figure 2a,b), while Rhod-DA displayed typical maximal excitation and emission wavelengths in methanol (560 and 585 nm, respectively) with a bright emission due to a high extinction coefficient (88200 M⁻¹ cm⁻¹) and a reasonable quantum yield of fluorescence (ϕ_F = 0.23).

Rhod-PM rapidly binds to 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC) large unilamellar vesicles (LUVs) (Figure S1) and, once embedded into this PM model, displayed slightly red-shifted spectra (Figure 2a) along with slightly reduced brightness compared to methanol (Figure 2b).

Self-Triggered Photooxidation Cascade in Bulk

To prove that a leuco-rhodamine can be photoactivated by a small amount of rhodamine following an STPC, we first assessed the ability of Rhod-DA to generate singlet oxygen as an oxidizing agent capable of oxidizing LRhod-DA. Since we expected the rhodamine to act as a photosensitizer, the quantum yields of singlet oxygen generation (ϕ_Δ) of Rhod-DA and Rhod-PM were evaluated through comparison with rose Bengal as a reference^[50] and by monitoring the fluorescence decay of DPBF (Figure 3a). The results of 0.96% and 0.69% (Figure 2b) were in line with the low intersystem crossing quantum yields of rhodamines (0.5% to 1.0%),^[51] and indicated that Rhod-PM could be used as a mild photosensitizer potentially able to oxidize LRhod-PM without being phototoxic to cells. Then, the ability of ¹O₂ to oxidize LRhod-DA was assessed by spectroscopy (Figure 3b). LRhod-DA was placed in a methanolic solution in the presence of ¹O₂ and the appearance of Rhod-DA was monitored by fluorescence spectroscopy. ¹O₂ was generated through the use of methylene blue, an efficient photosensitizer, which was independently excited (λ_Ex = 638 nm, ϕ_Δ = 0.50)^[50] and which provoked an increase in fluorescence intensity at 585 nm depicting the oxidation of LRhod-DA into its fluorescent oxidized form Rhod-DA. These experiments independently showed that Rhod-DA can be used as a photosensitizer to generate ¹O₂ and that the latter is able to oxidize LRhod-DA into its fluorescent Rhod-DA. Therefore, we assumed that Rhod-DA, by itself, could oxidize LRhod-DA. To check our assumption, a concentrated methanolic solution of LRhod-PM (10 µM) was irradiated at 532 nm by a laser beam over time. Upon irradiation, the fluorescence intensity integration increased for 2 h before slowly decreasing. The plotted curve corresponds to a typical autocatalysis sigmoid profile, relating to a first predominant photoactivation phase followed by a decrease assigned to a prevailing photobleaching phase (Figure 3c). This curve thus suggests 1) that Rhod-DA is generated over time and catalyzes the following oxidation of LRhod-DA and, 2) that upon its formation, Rhod-DA got photobleached. To verify our hypothesis, HPLC studies were conducted in parallel to monitor the photooxidation of LRhod-DA into Rhod-DA (Figure 3d). Before irradiation, both LRhod-DA (purple trace) and Rhod-DA (black trace) were analyzed by HPLC and revealed that LRhod-DA contained a small amount of the fluorescent Rhod-DA (≈5%, hashtag # in figure 3d, purple curve), probably due to oxidation over time by the oxygen contained in the air and solvents. Upon irradiation, the amount of LRhod-DA decreased (RT = 15.2 min), while Rhod-DA appeared (RT = 22.1 min) along with other photoproducts (e.g., RT = 21.2 min).

We also observed photoproducts that did not absorb in the visible range depicting photobleaching, whereas others were blue-shifted compared to Rhod-DA (peak indicated by Δ in figure 3d, λ_{Abs max} = 548 nm vs. 560 nm for Rhod-DA), corresponding to dealkylation upon irradiation.^[52] After calibration (Figure S2), we were able to determine a maximum chemical yield of oxidation from LRhod-DA to Rhod-DA of 53% after 3 h of irradiation.

Overall, these analyses confirmed the STPC process where upon irradiation a small amount of rhodamine acts as a photosensitizer capable of oxidizing leuco-rhodamine through autocatalysis followed by its photobleaching.

Self-Triggered Photooxidation Cascade of LRhod-PM in PM Model

After demonstrating that LRhod could be used as a photoactivatable probe, we assessed the STPC of LRhod-PM at different local concentrations. When LRhod-PM was irradiated in methanol at 200 nM, i.e., at a low local concentration, virtually no fluorescence enhancement was observed, probably due to the fact that the small number of Rhod-PM present in the bulk was too diluted to trigger the photooxidation cascade of LRhod-PM (Figure 4a). Conversely, when LRhod-PM was incorporated into the PM model: DOPC large unilamellar vesicles (LUVs) at a probe/lipid ratio of 1/250, a fluorescence enhancement at 589 nm of 38-fold was observed in 20 min, and a quantum yield of photoactivation was determined, ϕ_Pa = 5.8 × 10⁻⁵ (Figure 4b), which is in the range of photoactivatable fluorescent proteins.^{[53, 54]} This phenomenon was assigned to the increase in the local concentration of probes, thus favoring the oxidation of LRhod-PM by its parent photosensitizer Rhod-PM.

The same trend was observed for the photobleaching phenomenon when the fluorescent Rhod-PM was irradiated in methanol and in LUVs (Figure 4c,d). After 5 h of irradiation in methanol, a slight fluorescence increase was observed assigned to solvent evaporation while due to redistribution of the probe leading to a higher local concentration in LUVs, the kinetic of photobleaching was considerably increased compared to conditions in methanol when the probe homogeneously distributed, with a quantum yield of photobleaching, ϕ_Bl of 2.0 × 10⁻⁵ (Figure 4d). It was noteworthy that upon photobleaching in the LUVs a significant blue shift was observed, which is in line with our previous HPLC analysis and the photo-dealkylation process.^[52] Consequently, when localized in a lipid membrane, LRhod-PM alternatively undergoes enhanced photoactivation and enhanced photobleaching through an STPC due to increased local concentration. This hypothesis was further supported by the significant increase of both photoactivation and photobleaching kinetics upon the increase of local concentration through the variation of the probe/lipid ratio in the LUVs (1/100 to 1/1000) (Figure 4e). Indeed, the monitoring of the integrated signal over time combined with a kinetic model (see Supporting Information) allowed us to determine two kinetic constants k_Pa and k_Bl that were, respectively, assigned to the photoactivation and the photobleaching rates and yielded, respectively, the quantum yield of photoactivation ϕ_Pa and of photobleaching ϕ_Bl (Figure 4f). Interestingly, ϕ_Pa was found to be 7 to 10-fold higher than ϕ_Bl (Figure 4f). Overall, these experiments showed that Rhod-PM acts as a mild photosensitizer, able to oxidize LRhod-PM in an efficient manner when concentrated on a plasma membrane model through an STPC.

Photoactivation of LRhod-PM in Cell Imaging

Once the proof of concept was established, we assessed the STPC mechanism on cells using LRhod-PM. As controls, the plasma membrane staining of the probes was checked in HeLa cells using laser scanning confocal microscopy. Although LRhod-PM did not produce any signal, Rhod-PM displayed a fast and wash-free fluorescent plasma membrane staining with a high signal-to-noise ratio of 29 and a Pearson's colocalization coefficient of 0.80 with MemBright-Cy5.5 (Figure 5).

Then, cells incubated with LRhod-PM were irradiated for 40 s at 552 nm on a zoomed region at higher laser power (10%) and the cells were imaged over time (Figure 6a). Upon irradiation, the fluorescence signal progressively increased (5-fold) to reach a plateau after ≈30 s (Figure 6b, Supplementary movie 1). The obtained image displayed a good signal-to-noise ratio of 7 and indicated an efficient plasma membrane localization as proven by the high Pearson's colocalization coefficient with MemBright (Figure 6c).

Next, we confirmed that at the working concentration of 200 nM, both Rhod-PM and L-Rhod-PM were not cytotoxic nor photocytotoxic under irradiation using MTT cell viability assays in HeLa cells before and after irradiation (Figure S3). Overall, these cellular experiments successfully showed that L-Rhod can be used as an efficient photoactivatable plasma membrane probe without detrimental cytotoxic or photocytotoxic effects.

Self-Triggered Photooxidation Cascade Enables Live SMLM

Common PALM probes typically require UV photoactivation,^{[10, 11, 14, 55]} which is not preferable for live-cell applications. Conversely, LRhod-PM possesses an appealing feature for live-SMLM as it only requires a single laser in the visible range. Also, we assumed that the presented mechanism would find advantages in super-resolution imaging as it gathers the ideal conditions to perform live single-molecule localization microscopy (SMLM) in PALM conditions. Indeed, at the single-molecule level, an efficient photoactivation combined with an accelerated photobleaching would provide an efficient blinking profile and enable to reach a single-molecule regime where single emitters are temporally and spatially well separated.^[56] However, in live super-resolution imaging based on SMLM acquisition time should be low, i.e., the number of blinks should be relatively high to have an efficient reconstruction and to palliate the potential movement of the cell during imaging. HeLa cells were then stained with LRhod-PM at a low concentration of 20 nM. In widefield microscopy, the low signal-to-noise ratio virtually did not enable the discrimination of the plasma membrane. Consequently, the cells and focal plane were found using nuclear staining in an independent channel (Hoechst, λ_Ex: 405 nm). Upon irradiation of a region of interest (Figure 7a) at 532 nm (0.25 kW cm⁻²) a high frequency of blinking events was obtained and the single-molecule regime could be reached (See Supplementary movie 2). Importantly, the laser power density used here was significantly lower than many conventional SMLM experiments, typically 1 to 3 kW cm⁻².^[57] This blinking efficiency allowed the reconstruction of the plasma membrane from a 120 s acquisition only (9k frames at 14 ms frame rate) and with a high gain of resolution (Figure 7b). Indeed, while the widefield image after irradiation provided a membrane thickness of 681 nm, the SMLM image indicated 77 nm (Figure 7c). Due to the high brightness of Rhod-PM and its efficient excitation with the 532 nm laser line, the blinking event displayed a rather high mean localization precision of 21.1 ± 4.6 nm (Figure 7b). Moreover, the localized photooxidation of LRhod provided a fairly stable number of localizations along with a constant number of photons over time (Figure 7e,f).

In parallel, Rhod-PM was also evaluated for live SMLM. As expected, the same mean photon number per blink were obtained when using LRhod-PM and Rhod-PM since the former arises from the latter upon activation (Figure S4). Although reconstruction after acquisition provided an image of the plasma membrane (Figure S5), the use of Rhod-PM evidenced some drawbacks compared to LRhod-PM. Despite the use of poly-lysine-coated glass, Rhod-PM, as a cationic probe is prone to stick on the glass surface which induces a high background noise. This significantly reduces the localization density and the reconstructed image quality, which is depicted by the signal-to-noise ratio in widefield after irradiation (Figure S4B). Finally, while a cell stained with LRhod-PM reaches a single-molecule regime after only a few seconds, Rhod-PM, which is intrinsically fluorescent, needs to be photobleached for almost 1 min, which thus considerably reduces the reservoir of available blinking probes and increases the phototoxicity. This was illustrated by the mean event number of LRhod-PM which is twice higher than those of Rhod-PM over the acquisitions (Figure S4c, Supplementary movie 2). Overall, LRhod-PM enabled live SMLM imaging of plasma membrane in HeLa cells via STPC mechanism.

With LRhod-PM in hand, we then investigated the imaging of other PM protrusions and processes. Filopodia are finger-like projections that extend from the cell surface and play crucial roles in various cellular processes, including cell migration and signal sensing.^[58] Similarly, tunneling nanotubes (TNTs) are thin, tube-like PM processes that form transiently between cells, enabling direct intercellular communication.^[59] These PM structures are prone to movement and are thus challenging to image. However, using rapid acquisitions with LRhod-PM (1 min 24 and 28 s, respectively), we successfully imaged these thin processes in live HeLa cells with enhanced resolution (Figure 8a,b).

LRhod-PM was subsequently applied to two other cell lines to demonstrate its versatility: A549, human lung cancer epithelial cells (Figure 8c), and THP-1, human monocytes differentiated into macrophages (Figure 8d). Successive acquisitions were performed over 10 min on a TNT connecting two A549 cells. Owing to the performance of LRhod-PM, characterized by high blinking events over time and good localization precision despite the short acquisition time (14 ms), dynamic super-resolution imaging was achieved, resolving the two bilayers of the TNT (Figure 8c). Using the same approach, the movement and division of overlapping THP-1 cells were imaged at higher resolution, benefiting from the minimal fatigue associated with LRhod-PM's blinking performance (Figure 8d).