Genetically encoded dinitrobenzyl photostabilizer for enhancing the photostability of fluorescent proteins

Abstract

Fluorescent proteins (FPs) are essential tools for biological imaging, but their application in long-term and super-resolution microscopy is often limited by photobleaching. Inspired by “self-healing” organic dyes, we report a genetically encoded intramolecular photostabilization strategy to improve the ability of fluorescent proteins to resist photobleaching. We designed a dinitrobenzene-based unnatural amino acid (UAA), 2-NBAK, to function as an efficient photostabilizer. Using an engineered orthogonal translation system, we performed a systematic spatial scan of 192 insertion sites across the mNeonGreen (mNG) β-barrel. Our screening revealed that optimal mNG-2-NBAK variants at specific outward-pointing side chains are proposed to quench the triplet state without disrupting the chromophore emission spectra. In vitro characterization and further live-cell confocal imaging demonstrated that the optimal variants exhibited more than a twofold extension of the fluorescence half-life under intense continuous irradiation. By transplanting the optimal mNG insertion sites into another FP, we established a proof-of-concept for enhancing green FP photostability, providing a genetically encoded framework adaptable to other spectral variants with energetically matched photostabilizers.

We reported a genetically encoded custom “photostabilizer” amino acid to improve the ability of fluorescent protein (FP) to resist photobleaching.

Introduction

Genetically encoded fluorescent proteins (FPs) have revolutionized the field of biological imaging by providing non-invasive, highly specific tools with minimal physiological perturbation for visualizing biological processes and molecular sensing.^1–4 With the development of super-resolution microscopy and long-term confocal imaging, researchers can now observe the dynamic processes of subcellular structures at the nanoscale.^5–8 However, the high-intensity laser excitation required by these cutting-edge technologies places demands on the photophysical properties of fluorescent probes.^9,10 Under intense irradiation, FPs undergo irreversible photobleaching, which has become a core bottleneck for long-term structural and functional tracking.^11–13 Although directed evolution has successfully expanded the spectral range and imparted diverse functions, the improvement of photostability largely relies on random screening rather than rational design grounded in photophysical mechanisms.^14,15

Notably, while the three-dimensional β-barrel structures and the central chromophore frameworks of different FPs are highly conserved (Fig. 1A), their photophysical properties (e.g., brightness, emission spectra and photostability) often vary significantly.^16–19 This phenomenon strongly suggests that the photophysical characteristics of FPs depend not only on the chromophore itself, but also on the specific chemical microenvironment surrounding it within the barrel. Therefore, we propose that the photostability of FPs can be significantly improved by precisely adjusting the chemical microenvironment around the chromophore. Photobleaching of fluorescent probes typically occurs when the excited-state chromophore undergoes intersystem crossing (ISC) to enter a long-lived and reactive triplet state (T₁). The chromophore in the T₁ state is subsequently vulnerable to attack by reactive oxygen species (ROS) or free radicals in the local environment, leading to irreversible photobleaching.^20,21

Fig. 1. Rational design of photostabilizer unnatural amino acids and synthetase screening. (A) Structural alignment of the β-barrels and highly conserved chromophore frameworks of representative FPs (yellow: Hyperfolder YFP (PDB: 7UGR); green: mNeonGreen (PDB: 5LTR); blue: ECFP (PDB: 2WSN); red: mCherry (PDB: 2H5Q)). (B) Schematic illustration of the “intramolecular photostabilization” concept migration from self-healing dyes to FPs. (C) Photophysical processes of the chromophore and the photostabilization mechanism of the NBA moiety. Upon photon absorption, the S₀ chromophore transitions to the S₁ state and primarily returns to S₀via fluorescence emission. However, a fraction of the excited chromophores undergoes ISC to form a reactive, long-lived T₁. In the presence of ambient oxygen, the T₁ state can transfer energy to generate reactive singlet oxygen, which subsequently attacks the chromophore and causes irreversible photobleaching. The incorporated NBA moiety acts as an efficient quencher, rapidly depleting the T₁ state via an intramolecular redox cycle (ox/red) to provide a safe energy relaxation pathway back to S₀, thereby effectively suppressing photobleaching. (D) Chemical structures of the designed photostabilizer unnatural amino acids, 1-NBAK and 2-NBAK. (E) Screening of a MbPylRS mutant capable of recognizing both 1-NBAK and 2-NBAK. (F) Fluorescence characterization confirming the site-specific incorporation of 1-NBAK and 2-NBAK into the GFPuv-N149TAG reporter. Schematic diagrams were created with BioRender.com. FPs, fluorescent proteins; NBA, nitrobenzyl; S₀, ground state; S₁, first excited singlet state; T₁, triplet state; ISC, intersystem crossing; MbPylRS, Methanosarcina barkeri pyrrolysyl-tRNA synthetase.

To address this issue, several strategies have been developed, including the “self-healing dye” strategy.^22,23 This strategy involves covalent attachment of a triplet-state quencher or redox-active moiety to the fluorophore, thereby providing a safe relaxation pathway that returns the triplet-state chromophore to the ground state (S₀) and significantly delays photobleaching.^24,25 Inspired by this, we envisioned that covalently anchoring a photostabilizing moiety into the FP backbone could similarly modulate the chemical environment of the chromophore and achieve enhanced photostability (Fig. 1B).

Previously, Henrikus et al. utilized cysteine-maleimide chemistry to conjugate photostabilizers, such as azobenzene, to locally exposed sites (A206C and L221C) on the surface of the GFP-QC2 protein. Their results demonstrated a moderate improvement in photostability, preliminarily validating the feasibility of modulating FP photophysics via modifying groups.²⁶ However, traditional in vitro chemical modifications are usually restricted to a few accessible natural residues on the protein surface, making them difficult to use in live-cell imaging.^27,28 More importantly, the protective effect of a photostabilizer depends not only on its chemical nature, but also on its precise spatial distance and relative orientation with respect to the central chromophore.²⁹ Constrained by the scarcity of modifiable sites, traditional methods struggle to comprehensively elucidate the spatial distribution rules and orientational requirements underlying the photostabilization effect. Thus, there is an urgent need for a strategy capable of high-throughput, comprehensive site-specific insertion to systematically reveal the complex structure–activity relationships behind anti-photobleaching effects.

Herein, we employed genetic code expansion (GCE) to site-specifically incorporate photostabilizing moieties into FPs as unnatural amino acids. We designed two photostabilizer unnatural amino acids (PS-UAAs) based on nitrobenzyl (NBA), 1-NBAK and 2-NBAK, and conducted a systematic spatial scan across all 192 potential insertion sites located on the β-barrel of the bright fluorescent protein mNeonGreen (mNG). We successfully identified 67 modification sites where the incorporation of 2-NBAK yielded photostability comparable to or greater than the wild type (WT). In vitro photobleaching kinetics of purified proteins and time-lapse confocal imaging of live E. coli further confirmed that the optimal 2-NBAK variants (e.g., mNG-T170-2-NBAK and mNG-R182-2-NBAK) achieved more than a twofold enhancement in photostability. This study provides insights into the structure–activity relationships among modifier structure, spatial orientation, distance and FP photostability, providing a methodological reference for the future design of FP photophysical properties.

Results and discussion

Design and encoding of photostabilizer unnatural amino acids

Photobleaching significantly limits the application of fluorescent probes in long-term and super-resolution imaging. From a photophysical perspective, photobleaching primarily occurs when the excited-state chromophore undergoes ISC to enter a long-lived, highly reactive T₁. The chromophore in the T₁ state easily reacts with ambient oxygen molecules or amino acid residues in the β-barrel, generating destructive ROS or free radicals. This ultimately causes irreversible damage to the chromophore's conjugated structure (Fig. 1C).

To improve the photostability of fluorescent molecules, the “self-healing dye” strategy was developed by covalently linking specific groups to an organic chromophore as the photostabilizers. NBA is one of these photostabilizers and can effectively quench the reactive T₁ state through a rapid intramolecular redox reaction. This provides a safe energy relaxation pathway back to S₀, thereby significantly delaying photobleaching (Fig. 1B and C). Based on this mechanism, we transferred the “intramolecular photostabilization” strategy from organic dyes to genetically encoded fluorescent proteins.

To study the structure–activity relationships of FP photostability from a spatial topology perspective, we selected NBA as the core photostabilizing group. Because the spatial distance and relative orientation between the photostabilizer and the central chromophore are crucial for the anti-photobleaching effect, we designed and synthesized two PS-UAAs based on a lysine backbone: 1-NBAK and 2-NBAK (Fig. 1D). Specifically, the side chain of 1-NBAK contains a mononitro-substituted benzene structure, whereas 2-NBAK features a dinitrobenzene group. Furthermore, they differ in the length and structure of their linkers. This differential design provides structural diversity for the subsequent modification of mNG. The different linkers offer varied spatial conformations and orientational flexibility, while the differing number of nitro groups tunes the electronic properties and redox potential of the side chains. Together, these features enable systematic identification of the optimal topological positioning for the photostabilizing groups to exert their protective effects.

To efficiently and site-specifically introduce these bulky PS-UAAs into the target protein, we constructed an orthogonal translation system based on the pyrrolysyl-tRNA synthetase and its cognate tRNA from Methanosarcina barkeri (MbPylRS/tRNA^Pyl_CUA).³⁰ Using GFPuv-N149TAG as a reporter protein, we screened a previously constructed library of approximately 100 PylRS mutants. To facilitate the subsequent high-throughput screening of both UAA structures and protein sites, we deliberately selected a mutant from the positive clones that could simultaneously recognize both 1-NBAK and 2-NBAK. The final selected synthetase mutant, named NBAK-RS, contains the mutations Y271M/L274G/C313A (Fig. 1E and Table S1). Fluorescence characterization based on the GFPuv-N149TAG reporter system confirmed that this orthogonal system maintains a negligible background incorporation in E. coli and can efficiently direct the site-specific insertion of both 1-NBAK and 2-NBAK (Fig. 1F). This established a reliable tool for the subsequent comprehensive, high-throughput spatial scanning of the mNG β-barrel.

Systematic screening of photostabilizers and insertion sites for bright, photostable mNG

The location of a photostabilizer relative to the chromophore strongly influences its protective efficiency. However, introducing a bulky unnatural amino acid (UAA) into the tightly folded β-barrel can distort the surrounding microenvironment, often severely reducing or completely abolishing the intrinsic fluorescence of the protein. To find the optimal balance between effective photoprotection and high brightness, we systematically scanned 192 potential insertion sites across the 11 β-strands of mNG (Fig. 2A and B).

Fig. 2. Incorporation efficiency and structural tolerance of NBAK in mNG. (A) Schematic of the 192 selected candidate modification sites out of the 236 amino acid residues in the mNG β-barrel. (B) GCE-mediated NBAK incorporation and efficiency assay via fluorescence ratio (+/−UAA). (C) and (D) Incorporation efficiency and expression levels of 1-NBAK (C) and 2-NBAK (D) across 192 sites, yielding 105 and 87 bright variants, respectively, for subsequent photostability screening (see Fig. S1 and S3 for detailed residue positions). Schematic diagrams were created with BioRender.com. mNG, mNeonGreen; GCE, genetic code expansion.

Using the engineered orthogonal translation system, we first evaluated the incorporation efficiency of both 1-NBAK and 2-NBAK at these 192 sites in 96-well plates. To minimize background signal, we collected the cell pellets from the bacterial cultures and measured their fluorescence intensity in the presence or absence of the UAAs (for detailed experimental protocols, see Section S4). At this screening stage, we categorized the variants as “bright” or “dark” based on their whole-cell fluorescence, which reflects a combination of overall protein expression and proper chromophore maturation rather than intrinsic photophysical brightness. Specifically, a successful “bright” site should yield high absolute fluorescence (+UAA) to indicate good protein expression and proper chromophore maturation, alongside a high fluorescence ratio (+/−UAA) to confirm specific incorporation with low background mismatch. Based on these criteria, we identified 105 bright variants for 1-NBAK (+UAA fluorescence > 5000 and a +/−UAA ratio > 20) (Fig. 2C and Fig. S1). For 2-NBAK, 87 variants met the brightness threshold (+UAA fluorescence > 4000 and a +/−UAA ratio > 10) (Fig. 2D and Fig. S2A, B, S3). These results demonstrated that a significant portion of the mNG barrel could structurally tolerate these bulky NBAK insertions without disrupting the precise chemical environment required for light emission.

Having established a library of bright variants, we next sought to identify which UAA, at which insertion sites, could conferred resistance to photobleaching. We developed a live-cell, high-throughput photobleaching assay using E. coli cultures irradiated by a uniform 480 nm LED array. To ensure that the observed fluorescence decay was driven by photochemical bleaching rather than thermal denaturation caused by the light source, the multi-well plates were maintained in an ice bath during the continuous 30-minute irradiation (Fig. 3A, B and Fig. S4, S5). We continuously recorded the fluorescence retention rate of each variant to compare their photostability.

Fig. 3. Photostability screening of mNG-NBAK variants in live E. coli. (A) High-throughput photostability screening workflow for the bright variants via LED irradiation in an ice bath, with samples placed in the central position of wells to ensure uniform illumination. (B) Schematic of the in vivo photostability assay evaluating fluorescence retention, where higher retention under identical irradiation time indicates greater photostability. (C) and (D) In vivo photobleaching kinetics of the 105 mNG-1-NBAK (C) and 87 mNG-2-NBAK (D) bright variants expressed in E. coli. In both panels, curves for the WT and photostability-reduced variants are colored green and yellow, respectively. In (D), curves for the 67 favorable variants are colored light pink, with the top 12 variants highlighted in dark pink. (see Fig. S6 for detailed positions of the 67 favorable variants) (E) fluorescence retention rates at 30 minutes for the top 12 mNG-2-NBAK variants (dark pink) and the WT (green). (F) Spatial distribution of the top 12 photostable 2-NBAK modification sites mapped onto the mNG β-barrel (PDB: 5LTR). Schematic diagrams were created with BioRender.com. mNG, mNeonGreen; WT, wild type.

The screening results revealed a clear difference between the two photostabilizers. Among the 105 mNG-1-NBAK variants, none exhibited improved photostability compared to the WT; their fluorescence decayed at a similar or faster rate (categorized as “reduced” variants) (Fig. 3C). This indicates that the single nitro group, together with its particular linker in 1-NBAK, lacks the appropriate redox potential or spatial positioning to effectively quench the chromophore's triplet state. In contrast, the introduction of 2-NBAK yielded better outcomes. Out of the 87 bright mNG-2-NBAK variants, 67 demonstrated photostability comparable to or significantly greater than the WT (categorized as “favorable” variants), while the remaining 20 fell into the “reduced” category (Fig. 3D and Fig. S6). Furthermore, the degree of photoprotection varied greatly depending on the exact insertion site. This strong site-dependency supported our initial hypothesis: the protective effect depends strongly on the precise spatial distance and relative orientation of the photostabilizer with respect to the central chromophore.

Specifically, among these favorable candidates, the top 12 most photostable mNG-2-NBAK variants exhibited significantly enhanced fluorescence retention after 30 minutes of illumination compared to the WT (Fig. 3E). To gain mechanistic and structural insight into this topology-dependent protection, we mapped the structural distribution of the top 12 mNG-2-NBAK variants onto the crystal structure of mNG (Fig. 3F). Interestingly, the 3D structural mapping revealed that the side chains of 2-NBAK at these optimal positions predominantly pointed outward, away from the interior of the β-barrel. This specific orientation implies that the intramolecular photostabilization mechanism does not require direct, rigid steric contact with the internal chromophore. Instead, the necessary triplet-state quenching may occur efficiently without permanently altering the delicate internal packing, thereby maintaining the essential structural environment required for fluorescence emission and unaltered spectral properties.

In vitro characterization and live-cell confocal imaging of the optimal variants

To accurately evaluate the intrinsic photophysical properties of the variants without interference from complex cellular environments—such as varying protein expression levels or local redox state fluctuations—we purified the top 12 mNG-2-NBAK variants and the WT using standard affinity chromatography (Fig. 4A). We first measured their excitation and emission spectra. Maintaining the original spectral profile is crucial for a biological probe, as it ensures compatibility with existing microscope laser lines and filter sets. Notably, the spectra of all 12 variants were nearly identical to the WT (Fig. 4B). This structural and optical consistency support that introducing the bulky 2-NBAK onto the exterior of the β-barrel successfully preserved the delicate internal packing and the fundamental electronic states of the central chromophore.

Fig. 4. In vitro photostability validation and characterization of purified top 12 mNG-2-NBAK variants. (A) Schematic of the in vitro photostability assay for the purified protein variants, where higher retention under identical irradiation time indicates greater photostability. (B) Fluorescence spectra of the top 12 mNG-2-NBAK variants (yellow lines) and the WT (green lines), demonstrating preserved spectral properties without significant shifts. (C) In vitro photobleaching kinetics of five representative variants (2 µM) under continuous irradiation. The full dataset for the top 12 variants is shown in Fig. S7. The legend matches the top-to-bottom order of the curves. The control represents non-irradiated samples at the identical concentration. (D) Relative brightness and photostability of the top 12 variants. Relative brightness represents the initial fluorescence (t = 0) of purified proteins at 2 µM normalized to the WT (100%). The photobleaching half-life (t_1/2) indicates the time required for this fluorescence to decay by 50%. Schematic diagrams were created with BioRender.com. mNG, mNeonGreen; WT, wild type.

Next, we conducted rigorous in vitro photobleaching assays to quantify their absolute photostability and relative brightness independent of cellular expression levels (for detailed experimental protocols, see Section S10). Using the purified proteins at a matched concentration of 2 µM in an aqueous buffer, we subjected the samples to continuous LED irradiation. Under these standardized conditions, the WT mNG exhibited a rapid fluorescence decay with a half-life of just 8.5 min. In contrast, the most photostable variant, mNG-T170-2-NBAK, exhibited a half-life of 20 min, corresponding to a more than twofold improvement over the WT (Fig. 4C and Fig. S7). In practical probe development, researchers often face a trade-off between peak brightness and maximum photostability. To address different imaging needs, we plotted the relative brightness against the photostability for all candidates (Fig. 4D). From this distribution, we deliberately selected two outstanding variants for further live-cell validation: mNG-R182-2-NBAK (an optimal intermediate offering high brightness with meaningfully improved photostability) and mNG-T170-2-NBAK (the most stabilized candidate).

Finally, we tested these optimal variants under the high-intensity laser scanning conditions in live-cell confocal microscopy imaging (Fig. 5A). We expressed the selected variants and the WT in E. coli. To prevent bacterial motility from causing tracking errors or artificial fluorescence fluctuations, the live cells were carefully washed and immobilized on glass slides using 60% glycerol. We then performed time-lapse confocal imaging, continuously capturing images every 20 seconds for a total of 100 frames.

Fig. 5. Time-lapse confocal imaging of selected mNG-2-NBAK Variants in Live E. coli. (A) Schematic of the time-lapse confocal imaging assay. (B) Representative time-lapse fluorescence images of E. coli expressing the WT mNG and the two selected variants at 25-frame intervals. (C) Photobleaching kinetics of the corresponding variants in live E. coli under continuous confocal laser scanning. Schematic diagrams were created with BioRender.com. mNG, mNeonGreen; WT, wild type.

The imaging results qualitatively and quantitatively demonstrated the marked advantage of 2-NBAK insertion. Under the intense laser, the WT fluorescence signal dropped noticeably by frame 25 and was almost completely bleached by frame 50, resulting in a short half-life of approximately 25 frames. In stark contrast, cells expressing mNG-T170-2-NBAK and mNG-R182-2-NBAK maintained clear fluorescence signals even at the very end of the 100-frame recording (Fig. 5B). Quantitative analysis of 50 to 100 individual bacteria per image showed that the half-life of mNG-T170-2-NBAK exceeded 80 frames, while that of mNG-R182-2-NBAK was projected to exceed 120 frames (Fig. 5C). Taken together, these results demonstrate that the rational, site-specific insertion of 2-NBAK improves the photostability of mNG, which help to improve the photobleaching issues for long-term dynamic biological imaging.

Generalization of the intramolecular photostabilization strategy to sfGFP

To verify the cross-scaffold generalizability of our PS-UAA-based site-specific photostabilization strategy, we extended the system to another classic and widely used fluorescent protein, sfGFP. Given the highly conserved β-barrel topology between mNG and sfGFP, we directly mapped the top 12 photostable sites identified in mNG to their homologous positions in sfGFP via structural alignment. Structural analysis revealed that the overall side-chain trajectories and chromophore orientations at these mapped sites are highly conserved between the two scaffolds. Inspired by this high spatial similarity, seven of these equivalent sites (E34, N146, N170, V176, D180, Y182, and S202) in sfGFP were selected for targeted modification and preliminary in vivo screening in E. coli (Fig. S8A). All seven variants achieved successful site-specific incorporation of 2-NBAK with robust expression levels (Fig. S9). Subsequent photobleaching kinetics assays in live bacteria demonstrated that all seven variants exhibited varying degrees of photostability enhancement compared to the WT (Fig. S10). This indicates that with matched spatial topology, this intramolecular targeted quenching strategy holds potential to be applied across different chromophore microenvironments. Furthermore, similar to the observations in mNG, the incorporation of 2-NBAK barely altered the intrinsic fluorescence spectral properties of sfGFP (Fig. S11).

To more accurately evaluate their intrinsic photobleaching kinetics and mitigate the signal fluctuations inherent to the bacterial screening format, we selected five representative variants (sfGFP-E34, N146, N170, V176, and D180) for purified protein assays. Notably, three of these variants, N146, N170, and V176, demonstrated significant photostability improvements in vitro. Specifically, sfGFP-N146-2-NBAK and sfGFP-V176-2-NBAK exhibited near-zero fluorescence decay under continuous irradiation for 30 minutes, while sfGFP-N170-2-NBAK showed only a slight bleaching trend (Fig. S12). A detailed spatial comparison of the chromophore microenvironments for these three critical sites (Fig. S8B) provides crucial structural insights. Although the Cα atomic spatial deviations between sfGFP and mNG at these equivalent residue pairs vary (ranging from a minimal 0.3 Å for N170/T170 and V176/R176, to 2.3 Å for N146/D147), the native side chains at these positions consistently project toward the exterior rather than invading the internal core. More importantly, N146, N170, and V176 are all spatially clustered at the top portion of the β-barrel, specifically on the side facing the phenolic hydroxyl group of the chromophore. This spatial clustering suggests that the microenvironment near the phenolate moiety is a structural hotspot; positioning the quenching group here likely facilitates effective orbital overlap for highly efficient intramolecular triplet-state quenching across different scaffolds. However, despite this conserved topological hotspot, photoprotective efficiencies varied at equivalent positions between the two scaffolds. This suggests that beyond structural proximity, optimal quenching likely requires matching the specific triplet-state energies of the distinct chromophores (CRO in sfGFP vs. CR2 in mNG).

However, this enhanced photostability comes at a photophysical cost, highlighting a trade-off between stability and emission efficiency. To quantitatively evaluate this stability-brightness trade-off, we determined the fluorescence quantum yields of the engineered variants, as detailed in Table S2. A systematic decrease in quantum yield was observed across the top-performing photostabilized mutants compared to their WT counterparts. For instance, the values for the stable sfGFP variants (N146, N170, and V176) dropped significantly to between 0.11 and 0.18, which is well below the 0.65 value of wild-type sfGFP. A similar trend was observed in the most stable mNG variants, with the mNG-T170-2-NBAK variant dropping to a quantum yield of 0.13 from its wild-type value of 0.80. This consistent decrease in brightness and quantum yield across structurally equivalent positions may be attributed to two proposed mechanistic factors. First, the bulky side chain of 2-NBAK likely introduces significant steric hindrance near the chromophore. While the essential microenvironment for basic fluorescence is preserved, as evidenced by unaltered emission spectra, this steric crowding could slightly compromise the structural rigidity of the local β-barrel pocket. The potentially increased local flexibility might facilitate minor chromophore twisting or vibrations, which would in turn increase non-radiative decay pathways. Second, as our structural analysis revealed, these stable variants position the quenching moiety in close proximity to the phenolate hotspot. While this short distance effectively facilitates the desired triplet-state quenching, it may concurrently induce a degree of parasitic quenching of the singlet excited state.

The performance of these top variants reveal two key points. On the one hand, it strongly demonstrates that once the dinitrobenzene moiety is precisely positioned in an optimal spatial orientation, its intramolecular triplet-state quenching efficiency is sufficient to deliver near-complete photoprotection. On the other hand, it clearly highlights a limitation of the current strategy. Specifically, the consistent fluorescence attenuation observed at structurally equivalent positions across different scaffolds reveals that introducing bulky photostabilizing groups often entails a stability-brightness trade-off. Ultimately, these findings may guide the future design of next-generation PS-UAAs with smaller molecular volumes, enhanced spatial tolerance and tailored triplet-state energies.

Conclusions

In conclusion, we have developed a genetically encoded strategy to minimize photobleaching of fluorescent proteins by establishing a genetically encoded intramolecular photostabilization strategy. Using a rationally designed triplet-state quencher, 2-NBAK, and an engineered orthogonal translation system, we performed a spatial scan of 192 insertion sites across the mNG β-barrel. This comprehensive screening identified optimal variants, such as mNG-T170-2-NBAK and mNG-R182-2-NBAK, that demonstrate a more than twofold extension in fluorescence half-life under continuous irradiation, accompanied by a decrease in relative brightness. Furthermore, the emission spectra of the original chromophore are preserved. Structural analyses revealed that the primary shared topological feature among these top 12 mNG variants is an outward-facing 2-NBAK side chain. Given the highly conserved β-barrel topology between mNG and sfGFP, we mapped seven of top 12 mNG sites to their homologous positions in sfGFP. Among these, sfGFP-N146-2-NBAK, sfGFP-N170-2-NBAK, and sfGFP-V176-2-NBAK exhibited significantly enhanced photostability; notably, the corresponding mNG variants at these three equivalent positions were also among the most stable within the mNG top 12. We discovered that the most stabilized variants in both scaffolds consistently cluster near the phenolate moiety of the chromophore. This indicates that positioning an outward-facing quenching moiety at this specific structural hotspot enables efficient intramolecular protection without severely disrupting the delicate internal microenvironment.

Looking forward, our findings in mNG and sfGFP may serve as a structural blueprint for the site-specific intramolecular photostabilization of other fluorescent proteins, providing a foundation for developing stable, wash-free biosensors for live-cell imaging. To potentially address the observed stability-brightness trade-off and expand this strategy to other fluorescent scaffolds, future efforts could focus on identifying novel quenching moieties that introduce minimal steric perturbation while better matching the specific triplet-state energy of the target chromophore. Since evolving a specific aminoacyl-tRNA synthetase for every new quencher candidate is labor-intensive, a two-step workflow could be employed to accelerate this screening process. Specifically, an unnatural amino acid bearing a reactive handle could be incorporated at the identified spatial hotspots, followed by live-cell-compatible bioorthogonal chemistry (e.g., strain-promoted alkyne-azide cycloaddition (SPAAC) or inverse electron-demand Diels–Alder (IEDDA)) to rapidly screen diverse photostabilizing moieties. Ultimately, optimal candidates could be re-engineered into single-step GCE systems to provide the necessary topological rigidity and in situ expression. This efficient integration of tailored photophysics and precise genetic encoding will may facilitate the development of more photostable biosensors in advanced biological imaging.

Author contributions

G. L. and X. W. contributed equally to this work. J. W. conceived and supervised the project. G. L. and X. W. performed the experiments and analysed the data. J. X. conducted the chemical synthesis. All authors contributed to the writing and editing of the manuscript.

Conflicts of interest

There are no conflicts to declare.

Data availability

All experimental data supporting the findings of this study, including chemical synthesis procedures, NMR and MS spectra of the unnatural amino acid, protein sequences, in vitro spectroscopic characterizations, and imaging raw data, are available within the article and its supplementary information (SI).Supplementary information is available. See DOI: https://doi.org/10.1039/d6cb00168h.