Supercharged fluorescent proteins detect lanthanides via direct antennae signaling

Rare earth elements (lanthanides plus Sc and Y) are essential across sectors due to unique luminescent, magnetic, and catalytic properties. Current biosensing and recovery methods face challenges in detecting and handling lanthanides at concentrations found in environmental and industrial streams, which can reach ~20–100 µM in acid mine drainage and higher in process streams. Lanthanides are intrinsically luminescent, but 4f–4f transitions are Laporte-forbidden, limiting direct excitation. Sensitization via organic ligands (antenna effect) or lanthanide-to-fluorescent protein FRET can enable detection but typically requires tight chelation or engineered proximity, limiting scalability and in vivo use. Lanmodulin (LanM) provides picomolar sensitivity but may saturate in high-load streams. Observations that lanthanide cations bind negatively charged aspartate/glutamate residues on protein surfaces suggest an alternative approach. The study hypothesizes that protein charge engineering (supercharging) can create surface chelation sites that position lanthanides within nanometer-scale distances of fluorescent protein chromophores, enabling direct lanthanide-to-chromophore energy transfer upon UV excitation and quantitative detection across environmentally relevant ranges.

Prior work established the antenna effect where organic ligands sensitize lanthanide luminescence, and lanthanide-based FRET/LRET to fluorescent proteins when proximity is engineered. Lanmodulin (LanM) has picomolar affinity for lanthanides and has been used in fusion sensors such as LaMP1 (citrine–LanM–CFP) that report binding by FRET upon LanM folding. A tryptophan antenna strategy affords picomolar terbium detection but is limited to Tb and suffers from low quantum yield of tryptophan. While LanM-based sensors are ultra-sensitive, their operating range can be mismatched to higher concentrations typical of acid mine drainage and waste streams (tens of µM to mM), risking oversaturation. Reports of lanthanide binding to negatively charged protein surfaces (e.g., extracellular matrix) and advances in protein supercharging (mutating solvent-exposed residues to modulate net charge without disrupting fold) motivate exploring direct lanthanide chelation on fluorescent protein surfaces to drive energy transfer without specialized chelators.

Design and construction: The authors engineered supercharged variants of GFP, YFP, and CFP by mutating solvent-exposed residues to modulate net surface charge while preserving structure and chromophore environment. To red-shift/stabilize GFP into YFP variants, mutations T65G, V68L, S72A, and T203Y were introduced, yielding YFP variants with net charges −4, −10, −17, −31, +16, and +33 (pH 7.0). To blue-shift/stabilize GFP into CFP variants, mutations Y66W, F146G, N147I, H149D were introduced, yielding CFP variants with net charges −5, −11, −18, −32, +15, and +32 (pH 7.0). Together with the corresponding GFP series (−4, −10, −17, −31, +16, +33), 18 total biosensor variants were assembled.

Protein expression and purification: Synthetic genes (Twist Bioscience, IDT) were cloned in E. coli DH10B and expressed in BL21 using pET-based plasmids assembled by Golden Gate or supplied pre-assembled. Cultures were grown (LB, 37 °C), induced at OD600 ~0.6 with 1 mM IPTG after cooling, expressed 16 h at 18 °C, harvested, lysed by sonication, and purified by Ni-NTA (HisPur) with phosphate/imidazole buffers. Desalting and buffer exchange were performed (Amicon) into 50 mM Tris-HCl pH 7.0; proteins were concentrated to 0.4 mL, quantified by Bradford assay, and purity verified by SDS-PAGE.

Spectroscopy setup and definitions: Fluorescence measurements were recorded on a Cytation 5 reader at room temperature in 96-well plates, typically 0.1 mL volume, in 50 mM Tris-HCl pH 7.0. Lanthanide salts (TbCl3, TmCl3, DyCl3, EuCl3, SmCl3, YbCl3) and competitor salts (AlCl3, FeCl2, MgCl2, CaCl2, MnCl2, ZnCl2, CuCl2) were prepared in the same buffer. Protein concentration for assays was ~0.1 mg/mL (~3.7 µM). Time-resolved luminescence spectra were collected with detection delays of 0, 100, and 300 µs. An ‘excitation ratio’ metric quantified lanthanide-to-protein LRET: for GFP, emission at 510–520 nm upon excitation at 340 nm (antenna channel) divided by emission upon excitation at 465 nm (direct GFP channel); for YFP, emission at 530–540 nm upon excitation at 360 nm divided by emission upon excitation at 500 nm. ‘Factor change’ was defined as the excitation ratio with lanthanide divided by the biosensor-only ratio.

Screening and dose–response: The 18 variants were screened for increases in fluorescence upon UV excitation (250–400 nm) in the presence of Tb3+, Tm3+, Dy3+. Optimal ratiometric excitation was identified near 350–390 nm but to limit crosstalk, 340 nm (GFP) and 360 nm (YFP) were chosen; 280 nm excitation was avoided in quantitative analyses to exclude tryptophan antenna effects. Dose–response curves (1 nM–10 mM) were collected for Tb3+, Tm3+, Dy3+, and later Eu3+, Sm3+, Yb3+ with the top performers (GFP-10, YFP-31); apparent Kd values were estimated from response curves.

Time-resolved LRET validation: Long-lived luminescence consistent with lanthanide excited-state lifetimes was assessed for GFP-10 and YFP-31 in the presence/absence of 1 mM Tb3+ at delays of 0, 100, and 300 µs.

Interference and environmental robustness: Interference by Al3+ and Fe2+ was evaluated both as separate additions and in competition with a fixed 1 mM Tb3+. Broader interferents (Mg2+, Ca2+, Mn2+, Zn2+, Cu2+) were tested in competitive formats at up to 10-fold excess over Tb3+. Mixed-ion matrices approximating Virginia Canyon acid mine drainage composition (VC, containing Al3+, Fe2+, Mn2+, Zn2+, Cu2+; and mVC, Mn2+, Zn2+, Cu2+) were tested at pH 7 and pH 5. Ionic strength (0–800 mM NaCl), temperature (0–70 °C), and pH dependence were characterized. La3+ competition was tested up to 5-fold excess over Tb3+.

SuPrA supramolecular assembly: Using supercharged protein assembly, CFP+32 and GFP-31 were combined to form a globular protomer (including comparison to a known 16-unit CFP+32/GFP-17 assembly). FRET from CFP (ex 433 nm) to GFP was measured; effects of 1 mM Tb3+, Tm3+, Dy3+, or Al3+ on protomer FRET were assessed. UV excitation (340 nm) was then used to probe lanthanide antenna-mediated signaling within the assembly and quantify masked LRET signals by comparing fluorescence relative to the unperturbed protomer.

• Negatively supercharged YFP and GFP variants exhibited 2–3-fold increases in fluorescence upon UV excitation in the presence of Tb3+, Tm3+, or Dy3+, whereas positively charged variants showed minimal response. Yb3+, Eu3+, and Sm3+ also increased fluorescence with selected variants (GFP-10, YFP-31).
• Time-resolved spectroscopy showed long-lived luminescence consistent with lanthanide-mediated energy transfer: GFP-10 and YFP-31 paired with Tb3+ retained strong signals at 100 µs and 300 µs delays, while biosensors alone were negligible under delayed detection.
• Ratiometric performance statistics (factor change in excitation ratio, mean ± S.D.): • YFP series with Tb3+: YFP-31 = 1.97 ± 0.25 (highest); YFP-17 = 1.58 ± 0.25; YFP-10 = 1.48 ± 0.15; YFP-4 = 1.30 ± 0.17; YFP+16 = 1.13 ± 0.04; YFP+33 = 1.05 ± 0.03. • GFP series with Tb3+: GFP-10 = 2.58 ± 0.09 (highest among GFPs); GFP-31 = 1.60 ± 0.06; GFP-17 = 2.32 ± 0.25; GFP-4 = 1.82 ± 0.08; GFP+16 = 1.06 ± 0.24; GFP+33 = 0.93 ± 0.11.
• Sensitivity range: Top sensors YFP-31 and GFP-10 showed ~2.5-fold or greater increases in response to Tb3+, Tm3+, and Dy3+ across a dose-dependent 10 µM–5 mM range.
• Apparent Kd values across six lanthanides (Tb3+, Tm3+, Dy3+, Eu3+, Sm3+, Yb3+): YFP-31 ~25–30 µM; GFP-10 ~210–500 µM.
• CFP variants were non-performing across charges, attributed to unfavorable spectral overlap with Tb3+ and Tm3+ and excessive overlap with Dy3+.
• Interference testing: • Al3+ produced minimal perturbation; YFP-31 discriminated Tb3+ vs Al3+ with nearly 2-fold higher output for Tb3+ at 1 mM. • Fe2+ had little effect up to ~10 µM; higher concentrations reduced signal, with 1 mM Fe2+ weakening Tb3+ detection. • Competing divalent metals (Mg2+, Ca2+, Mn2+, Zn2+, Cu2+): minimal perturbation up to 10-fold excess; moderate effects at 1 mM Zn2+ and 100 µM Cu2+ (stronger for GFP-10 than YFP-31). No false positives without Tb3+. • Mixed ion matrices (VC and mVC) at pH 7 caused minimal perturbation even at 5-fold excess (VC). Equimolar mVC modestly interfered with GFP-10. At pH 5, detection was disrupted at ~10 µM (GFP-10:Tb3+) and ~1 mM (YFP-31:Tb3+) contaminant levels. • La3+ at 5-fold excess over Tb3+ did not significantly perturb Tb3+ detection.
• Environmental robustness: NaCl 0–800 mM minimally affected detection; temperature stability for Tb3+:GFP-10 was 0–70 °C and for Tb3+:YFP-31 was 0–60 °C; detection dropped below pH 5.
• SuPrA assembly (CFP+32/GFP-31 protomer): Lanthanides (Tb3+, Tm3+, Dy3+) disrupted CFP→GFP FRET more than Al3+ when excited at 433 nm, suggesting competitive binding/protomer effects. Under UV excitation (340 nm), partial restoration revealed masked LRET signals: Dy3+ factor 1.01 ± 0.09 (UV) vs 0.77 ± 0.04 (433 nm), Tm3+ 0.94 ± 0.05 vs 0.72 ± 0.11; Tb3+ showed a small, not significant UV increase, indicating assembly alters spectral overlap for Tb3+.

The study validates a direct lanthanide-to-chromophore LRET pathway enabled by protein surface charge engineering. Negatively supercharged GFP and YFP bind free lanthanide cations and position them within nanometer-scale distances to the chromophore, allowing UV excitation of the lanthanide antenna and efficient energy transfer to the fluorescent protein. This approach circumvents the need for dedicated organic chelators or complex fusion constructs and operates across the micromolar-to-millimolar concentration range typical of high-load environmental and industrial streams. Time-resolved measurements substantiate that the detected signal arises from lanthanide-mediated long-lived luminescence. The platform maintains functionality in the presence of common interferents (e.g., Al3+, mixed-ion matrices), with defined limits (e.g., high Fe2+), and is robust to ionic strength and temperature variations, though performance diminishes at pH < 5. The lack of response in CFP variants underscores the requirement for appropriate spectral overlap between lanthanide emission levels and fluorophore excitation bands. Extension to supramolecular assemblies demonstrates that higher-order architectures can interface with lanthanide antennae; although native FRET can be disrupted by lanthanide binding, UV excitation reveals masked LRET signals, suggesting opportunities to design materials that integrate both assembly and sensing. Overall, the findings address the need for scalable, genetically encodable, and environmentally relevant lanthanide sensors and point toward new paradigms for biosensing and materials-based REE processing.

This work introduces a simple, scalable sensing paradigm in which negatively supercharged fluorescent proteins directly chelate lanthanide cations and report their presence via lanthanide-to-chromophore energy transfer. Top performers (YFP-31 and GFP-10) quantitatively detect multiple lanthanides across 10 µM–5 mM, show favorable kinetics characteristic of lanthanide luminescence, and retain function in complex matrices and across broad ionic strength and temperature ranges. The approach expands the operational window beyond ultra-high-affinity LanM-based sensors, better matching high-load environmental and industrial scenarios. The concept generalizes to supramolecular protein assemblies, opening avenues for engineered materials that sense and capture REEs. Future directions include computational modeling to optimize surface charge patterns and chromophore environments, machine learning-guided design to tune affinity ranges and selectivity, integration into living systems for in vivo sensing and bioprocess monitoring, and development of macroscopic filtration or nanomaterial platforms leveraging this LRET modality.

• The sensing window primarily spans micromolar to millimolar concentrations; lower concentrations typical of pristine environments may require further engineering to increase affinity.
• Detection degrades at acidic pH (<5), limiting direct application to strongly acidic streams without preprocessing or buffer adjustments.
• High Fe2+ (≥10 µM, especially ~1 mM) and certain contaminants (e.g., Zn2+ at 1 mM, Cu2+ ≥100 µM) can attenuate signals; removal or mitigation (e.g., pH-controlled precipitation/oxidation of iron) may be necessary.
• CFP variants failed due to spectral mismatch, highlighting dependence on precise spectral overlap; sensor performance may vary with fluorophore selection and local environment.
• In supramolecular assemblies, lanthanides can disrupt native FRET, and LRET signals may be masked without appropriate excitation strategies; assembly state can shift spectral properties.
• Potential aggregation at near-equimolar high concentrations of oppositely charged species could affect readouts; careful handling and assay conditions are required.
• In vivo performance and long-term stability in real effluents remain to be demonstrated; complex matrices with >10-fold excess of multiple competing metals may necessitate further optimization.