Chemically stable fluorescent proteins for advanced microscopy

Fluorescent proteins (FPs) are essential tools for probing cell biology but often fail under conditions required for advanced microscopy, including chemical fixation, expansion microscopy (ExM), and correlative light and electron microscopy (CLEM). Many FPs lose fluorescence with aldehydes and osmium tetroxide (OsO4), and performance in hydrogel-based ExM is variable. The study aims to engineer fluorescent proteins with exceptional chemical and thermal stability, rapid folding/maturation, high brightness, minimal oligomerization, and compatibility with ExM and CLEM. Building on prior work with the bright, monomeric and relatively stable mGreen Lantern (mGL), the authors set out to create cysteine-free, chloride-insensitive, and chaotrope-resistant variants that could surpass superfolder GFP (sfGFP) in stability, function after harsh EM processing, and enable new applications such as fluorescence-assisted purification under denaturing conditions.

The field has evolved from the original GFP applications in cell biology to super-resolution modalities. Photoconvertible FPs like mEos4b and mEosEM have demonstrated resistance to aldehyde fixation and OsO4 for CLEM, but there remains a need for constitutively fluorescent GFP/YFP-like proteins with similar resilience. mGreen Lantern (mGL) was introduced as a bright monomeric GFP compatible with ExM and tissue clearing, with improved resistance to chemical and thermal denaturation compared with eGFP, mClover3 and mNeonGreen. Superfolder GFP (sfGFP) improved folding robustness but remains sensitive to strong chaotropes and some fixatives. Large Stokes shift (LSS) proteins such as mT-Sapphire and mAmetrine offer 405-nm excitation but suffer from cross-excitation or lower stability. This work addresses these gaps by engineering hyper-stable, cysteine-free variants and new LSS GFPs that avoid cross-excitation while maintaining high stability and performance.

• Protein engineering and structural biology: Structure-guided engineering from mGreen Lantern to create cysteine-free, chloride-insensitive variants. Crystal structures solved for hfYFP (PDB 7UGR, 1.7 Å), mhYFP (PDB 7UGS, 1.6 Å), and FOLD6 (PDB 7UGT, 1.2 Å) using molecular replacement from Clover templates. Mutations introduced to enhance stability (e.g., S147P, V206K, L195M) and monomericity.
• LSS FP engineering: Destabilized B-state chromophore by introducing T65S/Y203I (from mT-Sapphire) into hfYFP, followed by a targeted library (positions 203, 204, 205, 221, 222, 223 on β-strands 10/11) using degenerate codons (384 combinations). Screened colonies by eye for high 405-nm/low 470-nm excitation; selected variants were tested for GdnHCl stability in clarified lysate and purified protein. Identified LSSA12 (hfYFP-G65S/Y203I/Q204E/E222D/R223F) and further diversified to obtain LSSmGFP (hfYFP-T43S/G65S/L68Q/H77N/K140N/Y203I/V206K).
• In vitro stability assays: Kinetic unfolding in 6.3 M GdnHCl and 3.6 M GdnSCN at pH 7.5; equilibrium unfolding across GdnHCl/GdnSCN concentrations (24 h endpoints); isothermal melting at fixed temperatures (66.6–87.7 °C, up to 5 h) and temperature ramps (0.3 °C/min to 100 °C); H2O2 challenge (0–27% w/v, 15 min); pH titration and chloride sensitivity; NaOH denaturation kinetics (1 M NaOH) for select mutants.
• Cellular assays: Bacterial expression brightness and soluble/insoluble partitioning by SDS-PAGE densitometry; mammalian cellular brightness via P2A coexpression with mCherry in HeLa, HEK293T, BE(2)-M17; subcellular localization fusions (actin, tubulin, clathrin, ER, mitochondria, nuclei); oligomerization propensities via OSER assay; photostability under laser-scanning confocal.
• Fixative compatibility and ExM: Live/fixed fluorescence retention after 4% PFA or 4% PFA + 5% glutaraldehyde; performance in protein-retention ExM (proExM), including pre- and post-expansion fluorescence retention.
• CLEM workflows: OsO4 dose-response (0.001–1% for 1 h); time-course in 1% OsO4; cellular CLEM protocol with AAV-transduced HEK293 cells, aldehyde fixation (4% PFA/0.2% Glut), osmication (1% OsO4), agarose and OCT embedding, confocal imaging pre/post processing; HPF/freeze substitution and HM20 embedding with 100-nm section imaging.
• Fluorescence-assisted protein purification: Designed His6-FP-linker-TEV-POI fusions (POIs: mScarlet-I, Bacillus circulans xylanase, streptavidin). Performed IMAC under fully denaturing conditions (6 M GdnHCl) with visualization under 405/470-nm LEDs through filter goggles. Dialysis-based refolding, TEV cleavage, and native IMAC to recover POI; activity of streptavidin quantified by biotin-4-fluorescein quenching. Tested fluorescence persistence after overnight 6 M GdnHCl at 4 °C.
• Data analysis: Normalizations to native controls; exponential fits for kinetics; ANOVA with multiple comparisons for fixation/ExM retention; OSER scoring per established criteria.

• Hyper-stable YFPs: hfYFP (no cysteines, chloride-insensitive) exhibits exceptional stability. Spectroscopy (Table 1): hfYFP λex/λem 514/529 nm, Φ=0.60, ε=119,500 M−1 cm−1, molecular brightness 72, cellular brightness 2.6× eGFP, pKa 5.6, maturation 21 min, Tm 94.2 °C; weak dimer in OSER. mhYFP (monomer) λex/λem 515/529 nm, Φ=0.62, ε=124,000, brightness 77, cellular 2.5×, pKa 5.7, maturation 27 min, Tm 92.8 °C; monomer.
• Chaotrope and thermal stability: In 6.3 M GdnHCl, hfYFP becomes ~50% brighter and remains stable for at least 48 h; persisted >3 months in 7 M GdnHCl at RT, while sfGFP denatured immediately. In 3.6 M GdnSCN, hfYFP retained fluorescence for >40 min (half-life 9.3 min mono-exponential), outlasting others. Isothermal melting at 87 °C: t1/2 hfYFP 40.2 min; mGL 17.4 min; sfGFP t1/2 1.5 min; others denature within 8 min. Overall Tm: hfYFP 94.2 °C, ~20 °C above eYFP (72.9 °C) and ~14 °C above eGFP (80.3 °C).
• Chemical resilience: hfYFP withstood highest H2O2 concentrations among tested FPs; was more acid-stable (pKa 5.6) and chloride-insensitive compared to eYFP. hfYFP tolerated NaOH conditions (with S147P/V206K/L195M contributing to stability and monomericity in mhYFP).
• Cellular performance: In bacteria, ~80% of hfYFP was soluble vs ~50–60% for eGFP/mNG/eYFP/sfGFP; mGL 68%. In mammalian cells, hfYFP showed 33% higher brightness than eYFP and 2.4× vs eGFP; retained localization in common fusion targets. Photostability: hfYFP comparable to mGL but bleached faster than Clover and eYFP under confocal illumination.
• Fixative and ExM compatibility: After 4% PFA, hfYFP and eGFP lost ~20% fluorescence; mNG and eYFP retained 42% and 60%, respectively. After 4% PFA + 5% Glut, hfYFP retained 75% vs mGL 65%, eGFP 65%, mNG 29%. mhYFP compatible with proExM and retained 16% more fluorescence than eGFP post-expansion.
• CLEM robustness: Only mEos4b, hfYFP, and mhYFP tolerated CLEM-level OsO4. In 1% OsO4 for 1 h, hfYFP retained more fluorescence than mEos4b and mhYFP; all others were quenched within 10 min. In cellular CLEM processing (PFA/Glut, 1% OsO4, agarose/OCT), mhYFP cells retained ~35% of live-cell fluorescence—14× and 25× higher retention than mGL and eGFP, respectively.
• New LSS GFPs: LSSA12 (G65S/Y203I/Q204E/E222D/R223F) and LSSmGFP (T43S/G65S/L68Q/H77N/K140N/Y203I/V206K) are exclusively 405-nm-excitable (single excitation peak), with improved chemical/thermal stability. Table 1: LSSmGFP λex/λem 400/510 nm, Φ=0.48, ε=36,900, pKa 4.7, Tm 84.8 °C; LSSA12 λex/λem 398/511 nm, Φ=0.38, ε=38,700, pKa 4.6, Tm 93.9 °C. LSSmGFP’s photostability ~2× mAmetrine’s and comparable to mT-Sapphire, with reduced cross-excitation vs mT-Sapphire.
• Fluorescence-assisted purification: Using hfYFP or LSSmGFP fusions enabled visualization of all steps of denaturing Ni-NTA purification (6 M GdnHCl) with 405/470-nm LEDs. eGFP fusions denatured immediately upon solubilization; hfYFP/LSSmGFP remained fluorescent. Refolded streptavidin retained ~33% activity vs commercial. After overnight in 6 M GdnHCl at 4 °C, hfYFP fluorescence remained stable, whereas LSSmGFP and mGL were largely quenched.
• Solubility tag potential: hfYFP/LSSmGFP fusions increased soluble yield of diverse POIs versus eGFP fusions (e.g., mScarlet-1 soluble fraction +290% with hfYFP and +220% with LSSmGFP; Bcx soluble fraction +820% and +800%, respectively).

The engineered hyperfolder FPs address a key limitation of conventional GFP/YFPs by maintaining fluorescence under harsh chemical and thermal conditions encountered in advanced microscopy workflows. hfYFP’s extraordinary tolerance to chaotropes (including prolonged stability and even fluorescence enhancement in 6.3–7 M GdnHCl), high thermal stability, aldehyde and OsO4 compatibility, and chloride insensitivity make it particularly suited for CLEM and other challenging applications. mhYFP provides a more monomeric option with minimal trade-offs to stability and performance, benefiting difficult fusion contexts and ExM compatibility. The new LSS constructs (LSSmGFP, LSSA12) eliminate problematic cross-excitation seen with mT-Sapphire, while maintaining high stability and good photostability, enabling clean two-color imaging with GFP-like partners such as mGL. Structural insights from high-resolution crystal structures suggest that hydrophobic packing around the chromophore and stabilizing surface interactions may underlie resilience to denaturants, potentially allowing partial unfolding without loss of fluorescence. These properties, combined with tolerance to radical substitutions (e.g., elimination of conserved Trp/Cys), position hyperfolder FPs as robust scaffolds for biosensor engineering and directed evolution. The fluorescence-assisted purification approach leverages hfYFP/LSSmGFP stability to track denaturing IMAC workflows visually, and the data indicate potential roles as solubility tags that enhance expression and solubility of fusion partners. Overall, the findings show that no single FP fits every use case: mGL remains exceptionally bright in mammalian cells and may be preferable for some ExM tasks, while hfYFP excels in CLEM due to OsO4 tolerance comparable to mEos4b. The portfolio of hyperfolder and LSS variants broadens the toolset for both routine and specialized imaging and purification applications.

This work introduces hyperfolder YFPs (hfYFP and mhYFP) with unprecedented chemical and thermal stability, compatibility with aldehyde and osmium fixation, and robust cellular performance, as well as two large Stokes shift GFPs (LSSmGFP, LSSA12) that avoid cross-excitation artifacts while maintaining high stability. High-resolution structures provide templates for further engineering and sensor development. The hyperfolder FPs enable new practical applications such as fluorescence-assisted purification under denaturing conditions and may act as solubility tags to improve yields of difficult proteins. Future research directions include mechanistic studies (e.g., molecular dynamics and free energy calculations) to elucidate the basis of chaotrope and alkali resistance, transferring stabilizing features to other FP families, evaluating intracellular metabolic fate and turnover of hyper-stable FPs, and expanding the suite of biosensors and colors using these scaffolds.

• Cellular fate and turnover: The impact of high stability on metabolic degradation and potential lysosomal accumulation is unknown and warrants verification in specific fusion contexts and turnover assays.
• Photostability considerations: hfYFP bleaches faster than Clover or eYFP under confocal illumination; prolonged high-intensity imaging may require caution or alternative FPs.
• Denaturing storage stability: LSSmGFP (and mGL) lose substantial fluorescence after overnight storage in 6 M GdnHCl at 4 °C, making hfYFP preferable for extended denaturing workflows.
• Generalizability of refolding: Not all POIs refold efficiently after denaturing purification under the reported conditions (e.g., Bcx did not), requiring case-by-case optimization of refolding buffers and protease compatibility.
• Oligomerization propensity: hfYFP scores as a weak dimer in OSER (though performs well in fusions); mhYFP addresses monomericity with slightly reduced stability.
• Application-specific trade-offs: While hfYFP excels in CLEM, other FPs (e.g., mGL) may be brighter in mammalian cells for applications like ExM; selection should be tailored to the modality.