De Novo Luciferases Enable Multiplexed Bioluminescence Imaging

Artificial Intelligence Information Science Medical Imaging Bioengineering Biochemistry Oncology Life Sciences Medicine
de novo luciferases multiplexed imaging bioluminescence protein design AI design

De novo AI-designed luciferases enable multiplexed bioluminescence imaging

Academic Background

Bioluminescence technology is a highly sensitive and non-invasive imaging technique, enabling real-time monitoring in living organisms without the need for external light sources. Luciferase is the key enzyme that catalyzes the light-emitting reaction, but natural luciferases face multiple limitations, such as poor protein folding, large size, ATP dependence, and low catalytic efficiency. These limitations have hampered the widespread application of bioluminescence technology in biomedical research. In recent years, although some progress has been achieved by modifying native luciferases through methods such as directed evolution, these approaches have not fully overcome the inherent limitations.

To address these challenges, the research team utilized deep-learning-based protein design methodologies to de novo design a novel class of luciferases, named the NeoLux series. These artificially designed luciferases exhibit excellent catalytic efficiency, stability, small size, ATP independence, and can also be fused with fluorescent proteins (FP) to generate highly efficient Förster Resonance Energy Transfer (FRET) systems, enabling multiplexed bioluminescence imaging. This research opens up new prospects for the application of bioluminescence technology.

Source of the Paper

This study was conducted by Julie Yi-Hsuan Chen, Qing Shi, Xue Peng, and others from the University of California, Santa Cruz. The corresponding author is Andy Hsien-Wei Yeh. The paper was published in Chem on March 13, 2025, under the title “De Novo Luciferases Enable Multiplexed Bioluminescence Imaging”.

Research Process and Results

1. Computational Design and Experimental Validation of Second-generation Luciferases

The research team first used the deep learning model ProteinMPNN to explore the sequence space of the first-generation luciferase LuxSit-I, designing 10,000 new amino acid sequences. Structural predictions of these sequences were conducted using AlphaFold2, and 191 candidate sequences were selected for experimental validation. Among them, 134 sequences showed luciferase activity, indicating that the designed sequences could retain their original structure and function. Further selection yielded the 20 designs with the highest activity for large-scale expression and purification, ultimately resulting in 16 monomeric luciferases.

To further enhance luciferase activity, the team conducted combinatorial mutagenesis screening on the catalytic pocket of NeoLux1, obtaining NeoLux1.2 with a 47% increase in activity. Experiments showed that NeoLux1.2 possesses extremely high thermostability (Tm > 100°C) and high specificity for the synthetic substrate DTZ (diphenylterazine). Additionally, NeoLux1.2 has a remarkably long signal decay half-life of 43 minutes, which is much longer than natural luciferases, making it more suitable for high-throughput screening.

2. Design of Luciferase-Fluorescent Protein FRET Systems

To achieve multiplexed bioluminescence imaging, the research team fused NeoLux1.2 with a variety of fluorescent proteins (such as mNeonGreen, mGold, mKok, CyOFP1, mKate2) to construct efficient FRET systems. They used AlphaFold2 to predict the structures of the FRET pairs, optimizing the distance between luciferase and fluorescent proteins to improve energy transfer efficiency. Five FRET pairs (LuxNeon, LuxGold, LuxKok, LuxOFP, LuxKate) were experimentally validated, all achieving energy transfer efficiencies greater than 90% and retaining the spectral properties of the respective fluorescent proteins.

3. Applications of Multiplexed Bioluminescence Imaging

The team validated the multiplexed imaging capability of these FRET systems in both cells and live mice. In cellular experiments, signals from different FRET pairs were successfully distinguished using linear unmixing techniques, enabling simultaneous imaging of multiple subcellular structures. In mouse experiments, HeLa cells expressing different FRET pairs were transplanted into mice, and following intravenous injection of the DTZ substrate, multiplexed bioluminescence imaging was successfully achieved. Moreover, the researchers demonstrated that these FRET systems allow real-time monitoring of dynamic changes in tumor heterogeneity, showing potential value in cancer research.

Research Conclusions and Significance

This study, driven by artificial intelligence-based protein design, successfully developed a new class of luciferases—the NeoLux series. These luciferases not only overcome many of the limitations of native luciferases, but also enable highly efficient multiplexed bioluminescence imaging through fusion with fluorescent proteins. The research team demonstrated the broad applications of these FRET systems in both cellular and in vivo settings, including multicolor subcellular imaging and real-time monitoring of tumor heterogeneity.

This research opens up new perspectives for bioluminescence technology, particularly in cancer research, drug screening, and biomedical imaging. Furthermore, the team provides open access to the design sequences and experimental protocols, significantly reducing experimental costs and promoting broad application of this technology in laboratories.

Research Highlights

  1. AI-driven protein design: Using deep learning models ProteinMPNN and AlphaFold2, the team de novo designed novel luciferases, showcasing the powerful role of AI in protein engineering.
  2. Multiplexed bioluminescence imaging: By constructing efficient FRET systems, the study achieved multiplexed bioluminescence imaging in cells and in vivo, offering new tools for investigating complex biological processes.
  3. High stability and specificity: The NeoLux series luciferases feature exceptional thermostability and substrate specificity, overcoming the limitations of natural luciferases.
  4. Low cost and wide applicability: The team provides open-access design sequences and experimental protocols, significantly lowering experimental costs and promoting the widespread adoption of this technology.

Other Valuable Information

The research team also demonstrated the use of these FRET systems for real-time monitoring of tumor heterogeneity, providing new tools for cancer studies. Additionally, they developed cost-effective multiplexed bioluminescence imaging methods, offering economical solutions for biomedical research.

This research not only advances bioluminescence technology but also provides a new paradigm for the application of artificial intelligence in protein design, with significant scientific and practical value.