Periodic Reporting for period 2 - CC-LEGO (CC-LEGO: robust protein blocks to build cages and layers)
Reporting period: 2020-09-01 to 2021-08-31
The design of new proteins offers significant advantages for human health, such as improved vaccines, drug delivery and novel therapies (e.g. a COVID-19 inhibitor, flu vaccine, treatment of Alzheimer’s). Proteins made of a single polypeptide chain that can take a wide variety of useful shapes can be reliably designed. However, for drug delivery and vaccine applications, larger protein assemblies (including protein cages) are needed.
Designing large protein assemblies is difficult, with success rates below ten percent, since atomically accurate interfaces between two protein chains must be designed. The process involves correctly predicting many complex chemical interactions (hydrogen bonding, hydrophobic effect, Van-der-Walls forces, and more).
The main aim of the project is to enable easier design of protein cages by circumventing the interface design problem. This is achieved by using existing designed Coiled-Coils (CC) at the cage interface. CCs are some of the best understood protein structures and can be designed with high success rates (>50%). Using sophisticated molecular design software, these CC structures can be rigidly fused to other proteins to form a set of LEGO-like building blocks that can be used to reliably make cages as well as other useful protein nanostructures.
A secondary objective is the transfer of knowledge and protein design know-how from a best-in-class protein design institution (the Baker Lab, Washington, USA) to the European Union (National institute of Chemistry, Ljubljana, Slovenia).
In conclusion, both objectives have been achieved. We have created pH responsive cages made from rigidly fused building blocks. The Experienced Researcher (Dr. Ljubetič) has learned the computational and experimental side of de novo protein design and has brought the skills and knowledge to the European Union (National institute of Chemistry, Ljubljana, Slovenia), where he is starting up his own group. His career has received a large boost and he has obtained Slovenian funding for his further research.
Designing large protein assemblies is difficult, with success rates below ten percent, since atomically accurate interfaces between two protein chains must be designed. The process involves correctly predicting many complex chemical interactions (hydrogen bonding, hydrophobic effect, Van-der-Walls forces, and more).
The main aim of the project is to enable easier design of protein cages by circumventing the interface design problem. This is achieved by using existing designed Coiled-Coils (CC) at the cage interface. CCs are some of the best understood protein structures and can be designed with high success rates (>50%). Using sophisticated molecular design software, these CC structures can be rigidly fused to other proteins to form a set of LEGO-like building blocks that can be used to reliably make cages as well as other useful protein nanostructures.
A secondary objective is the transfer of knowledge and protein design know-how from a best-in-class protein design institution (the Baker Lab, Washington, USA) to the European Union (National institute of Chemistry, Ljubljana, Slovenia).
In conclusion, both objectives have been achieved. We have created pH responsive cages made from rigidly fused building blocks. The Experienced Researcher (Dr. Ljubetič) has learned the computational and experimental side of de novo protein design and has brought the skills and knowledge to the European Union (National institute of Chemistry, Ljubljana, Slovenia), where he is starting up his own group. His career has received a large boost and he has obtained Slovenian funding for his further research.
Initial testing of CC interfaces designed in Chen et. al (Nature, 2019) revealed that certain designed interfaces can become stuck together in unfavorable configurations, preventing the formation of more desirable nanostructures. No binding occurs when these same protein chains were prepared separately and then mixed. Therefore, we focused on designing kinetically reversible heterodimers, that bind in the desirable conformation. We developed a computation pipeline to choose more polar interfaces (27336 interfaces were screened) and experimentally tested the binding of computationally chosen designs using a high throughput split-luciferase assay (50 constructs were tested). We obtained a reversible CC named mALb8.
We rigidly fused CCs to other designed proteins (Brunette et. al, Nature 2015) to form the CC-LEGO blocks. We computationally screened ~490.000 possible arrangements. Due to the expected higher success ratio, most of the CC-Lego blocks were only tested in the context of the higher order assemblies, not individually.
The CC-LEGO blocks were designed into cages using the novel WORMS methodology (Hsia & Mout, Nature Communications, 2021). A promising cage (I05-37) was tested by Dr. Joshua Lubner (Baker Lab). The electron micrograph imaging (and it’s 3D reconstruction) shows an excellent match with the design structure.
We have also focused on developing larger scale structures. Using the rigid fusion methodology, we have created fibres that span several micrometres in length. We have solved the structure of one of the fibre using Cryo-EM. This would not have been possible without the rigidly attached designed repeat proteins that served as markers for the single particle reconstitution. Additionally, we have demonstrated a practical application of the fibers, by attaching heterodimeric binders.
The work has so far been published in four scientific papers (an * indicates shared first or corresponding authorship):
• KOEHLER LEMAN, LYSKOV, LEWIS, …, LJUBETIČ, …, GRAY, BONNEAU. Ensuring scientific reproducibility in bio-macromolecular modeling via extensive, automated benchmarks. Nature communications. 29 Nov. 2021
• HUNT, …, LJUBETIČ, …, VEESLER, JEWETT, BAKER. Multivalent designed proteins neutralize SARS-CoV-2 variants of concern and confer protection against infection in mice. Science translational medicine, 2022
• LINDER, LA FLEUR, CHEN, LJUBETIČ, BAKER, KANNAN, SEELIG. Interpreting neural networks for biological sequences by learning stochastic masks. Nature machine intelligence, 2022
• DAVE, MASSARANO, KATZIR, STRMŠEK, LJUBETIČ*, SEMENTA*. EMBO beyond biology: connecting peptide, protein, and DNA design with systems chemistry. Chem, 2022.
At least two further papers are in preparation.
The work has so far been presented at 12 scientific conferences (7 lectures and 5 posters).
Several outreach events were organized, including 4 radio interviews and a Rosetta Workshop “De novo design of proteins using Rosetta and Alphafold 2” (https://sites.google.com/view/rosettacrashcourse) with over 60 applicants, proving that there is a lot of interest in protein design at the National Institute of Chemistry (Slovenia) and wider region.
Knowledge transfer to the EU has been successful, for example dr. Ljubetič has installed Rosetta, Alphafold2, ProteinMPNN and the other deep learning software on the computing cluster of National Institute of Chemistry and made them available to all researchers there.
We rigidly fused CCs to other designed proteins (Brunette et. al, Nature 2015) to form the CC-LEGO blocks. We computationally screened ~490.000 possible arrangements. Due to the expected higher success ratio, most of the CC-Lego blocks were only tested in the context of the higher order assemblies, not individually.
The CC-LEGO blocks were designed into cages using the novel WORMS methodology (Hsia & Mout, Nature Communications, 2021). A promising cage (I05-37) was tested by Dr. Joshua Lubner (Baker Lab). The electron micrograph imaging (and it’s 3D reconstruction) shows an excellent match with the design structure.
We have also focused on developing larger scale structures. Using the rigid fusion methodology, we have created fibres that span several micrometres in length. We have solved the structure of one of the fibre using Cryo-EM. This would not have been possible without the rigidly attached designed repeat proteins that served as markers for the single particle reconstitution. Additionally, we have demonstrated a practical application of the fibers, by attaching heterodimeric binders.
The work has so far been published in four scientific papers (an * indicates shared first or corresponding authorship):
• KOEHLER LEMAN, LYSKOV, LEWIS, …, LJUBETIČ, …, GRAY, BONNEAU. Ensuring scientific reproducibility in bio-macromolecular modeling via extensive, automated benchmarks. Nature communications. 29 Nov. 2021
• HUNT, …, LJUBETIČ, …, VEESLER, JEWETT, BAKER. Multivalent designed proteins neutralize SARS-CoV-2 variants of concern and confer protection against infection in mice. Science translational medicine, 2022
• LINDER, LA FLEUR, CHEN, LJUBETIČ, BAKER, KANNAN, SEELIG. Interpreting neural networks for biological sequences by learning stochastic masks. Nature machine intelligence, 2022
• DAVE, MASSARANO, KATZIR, STRMŠEK, LJUBETIČ*, SEMENTA*. EMBO beyond biology: connecting peptide, protein, and DNA design with systems chemistry. Chem, 2022.
At least two further papers are in preparation.
The work has so far been presented at 12 scientific conferences (7 lectures and 5 posters).
Several outreach events were organized, including 4 radio interviews and a Rosetta Workshop “De novo design of proteins using Rosetta and Alphafold 2” (https://sites.google.com/view/rosettacrashcourse) with over 60 applicants, proving that there is a lot of interest in protein design at the National Institute of Chemistry (Slovenia) and wider region.
Knowledge transfer to the EU has been successful, for example dr. Ljubetič has installed Rosetta, Alphafold2, ProteinMPNN and the other deep learning software on the computing cluster of National Institute of Chemistry and made them available to all researchers there.
The CC-LEGO blocks will be a useful building block for many higher order assemblies, such as cyclic oligomers, cages, fibers and 2D layers. The WORMS method of finding rigid fusions that satisfy geometric criteria is in theory able to sample such higher order assemblies.
The kinetically reversible mALb8 interface has many advantages over existing de-novo interfaces. It enables a mix-and-match approach to building two-component assemblies. It is also advantageous for designing additional protein cages, since weaker interfaces result in more cages assembling fully (if an interface is too strong, partially assembled kinetically trapped intermediates are probable).
pH responsive cages have countless applications in endosomal drug delivery. A single protein cage can contain large amounts of cargo (e.g. a cancer-fighting drug) that can be released upon intake into target cells. They can also form the basis of novel vaccines.
Fibers (large 1D constructs) with rigid fusions have many potential applications. The fusions can serve as markers that can enable Cryo-EM structural determination. Heme binding proteins could be fused to the fibers to create nano-wires that can conduct electricity. The fibers with heterodimeric binders can also serve as scaffolds for more complicated nano-machines.
The knowledge and experience that dr. Ljubetič has gained in the Baker lab (Seattle, USA), a world-renowned laboratory for protein design, has enabled him to start his own group in the Department of Synthetic Biology and Immunology at the National institute of Chemistry (Ljubljana, Slovenia). This will further advance de-novo protein design in the region.
In summary, the action has helped advance the state-of-the-art of protein cage and rigid fiber design for various applications, including human health.
The kinetically reversible mALb8 interface has many advantages over existing de-novo interfaces. It enables a mix-and-match approach to building two-component assemblies. It is also advantageous for designing additional protein cages, since weaker interfaces result in more cages assembling fully (if an interface is too strong, partially assembled kinetically trapped intermediates are probable).
pH responsive cages have countless applications in endosomal drug delivery. A single protein cage can contain large amounts of cargo (e.g. a cancer-fighting drug) that can be released upon intake into target cells. They can also form the basis of novel vaccines.
Fibers (large 1D constructs) with rigid fusions have many potential applications. The fusions can serve as markers that can enable Cryo-EM structural determination. Heme binding proteins could be fused to the fibers to create nano-wires that can conduct electricity. The fibers with heterodimeric binders can also serve as scaffolds for more complicated nano-machines.
The knowledge and experience that dr. Ljubetič has gained in the Baker lab (Seattle, USA), a world-renowned laboratory for protein design, has enabled him to start his own group in the Department of Synthetic Biology and Immunology at the National institute of Chemistry (Ljubljana, Slovenia). This will further advance de-novo protein design in the region.
In summary, the action has helped advance the state-of-the-art of protein cage and rigid fiber design for various applications, including human health.