BBa_25BR1T4B
- Type: Coding
- Short description: Nuclear-targeted anti-GFP nanobody
- Designer: Kristina Boneß
- Length: 209 amino acids
- Description from registry: This construct encodes an anti-GFP nanobody with an N-terminal pelB signal peptide for periplasmic localization and an 5x aspartate linker between pelB and nanobody coding sequence for subsequent extracellular secretion. The 5x aspartate linker is followed by a HRV 3C protease cleavage site at its C-terminus to cleave off the linker. At the C-terminus of the nanobody coding sequence a 3x nuclear localization sequence (NLS) is supposed to direct the protein into the nucleus of eucaryotic cells. C-terminally of the 3x NLS is a TEV protease cleavage site to cleave off the C-terminal 6x His-tag, which is used for protein isolation and purification. This construct was synthesized for us and cloned into pET blank vector with RBS and a resistance gene against kanamycin.
Achievements
Important note: We did not manage to express this particular construct experimentally due to time issues. However, we expressed a similar construct, RBSp_aGFP_his, which contains different restriction enzyme recognition sites not compatible with the iGEM Parts Registry.
BBa_2503U531
- Type: Coding
- Short description: Nuclear-targeted GFP (basic)
- Designer: Kristina Boneß
- Length: 309 amino acids
- Description from registry: This construct encodes the green fluorescent protein (GFP) with an N-terminal pelB signal peptide for periplasmic localization in E. coli. At the C-terminus of the GFP coding sequence a 3x nuclear localization sequence (NLS) is supposed to direct the protein into the nucleus of eucaryotic cells. C-terminally of the 3x NLS is 6x His-tag, which is used for protein isolation and purification. This construct was synthesized for us and cloned into pET blank vector with RBS and a resistance gene against kanamycin.
Achievements
Important note: We did not manage to express this particular construct experimentally due to time issues.
BBa_257TMLA6
- Type: Coding
- Short description: Nuclear-targeted anti-GFP nanobody (basic)
- Designer: Kristina Boneß
- Length: 186 amino acids
- Description from registry: This construct encodes an anti-GFP nanobody with an N-terminal pelB signal peptide for periplasmic localization in E. coli. At the C-terminus of the nanobody coding sequence a 3x nuclear localization sequence (NLS) is supposed to direct the protein into the nucleus of eucaryotic cells. C-terminally of the 3x NLS is 6x His-tag, which is used for protein isolation and purification. This construct was synthesized for us and cloned into pET blank vector with RBS and a resistance gene against kanamycin.
Achievements
Important note: We did not manage to express this particular construct experimentally due to time issues.
BBa_25OWWAE9
- Type: Coding
- Short description: In silico designed Alpha-Amanitin binding protein
- Designer: Milo Krömer
- Length: 103 amino acids
-
Description from registry: We did AI-based protein design to create a protein binding tightly to alpha-amanitin, the bi-cyclic peptide toxin of the death cap mushroom.
We set out to computationally design a nanobody with the same function but soon had to realize that our peptide epitope was too small and too modified by posttranslational modifications (PTMs) to be usable by the nanobody design pipeline we employed. With expert feedback, we decided to create non-immunoglobulin proteins in the size range of nanobodies that better accommodate the binding of our target.
We chose AI-based models that were recently released and are specialized on small molecule binder generation capable of considering all atoms not just amino acid residues. The model we chose was Boltzdesign1, an inverted neural network for structure prediction, optimized to generate structures for small molecule, DNA/RNA and PTM binding proteins.
With this model, we created 50 designs that underwent an iterative process of selection, cross-validation and parameter prediction to find the most feasible design.
Initial selection with internal quality metrics set 14 designs apart from the rest as high-quality ones. The structure and ligand binding of these was re-predicted with Alphafold3 and Boltz2 to find homogeneous predictions as cross-model-validation.
The high-quality designs were assessed for different protein characteristics with classical bioinformatic tools and neural networks predicting solubility, usability, affinity and propensity to aggregate.
From all the data, one design candidate stood out during quality comparison as well as trait prediction.
Figure 1: Overall protein fold and ligand positioning of design #15 Design #15 performed exceptionally well in internal quality metrics, especially in the confidence scores regarding the ligand positioning and binding prediction. It also had the best predicted distance error score (ipde) for the ligand/protein interface. Additionally, the root mean square deviation (RMSD) between the design output and each following repeated predictions from sequence with different models was below 2Å, signifying desired structure similarity. The models produced very homogeneous structure outputs and even the ligand was predicted in the same conformation and binding mode each and every time with a ligand RMSD below 2Å.
Figure 2: Overall ligand conformation homogeneity from 20 different Boltz2 predictions, superimposed to the original Boltzdesign1 output complex of design #15 This gave us confidence that the design and binding mode was valid. In predictions of solubility, usability, aggregation propensity and affinity, the design #15 performed among the top ranks of all 50 designs (for an extended comparison of #15 to other designs see our Drylab and modelling page here)
Design #15 did show many binding specifics common in strongly interacting protein/ligand structures, like hydrogen-bonding, hydrophobic interactions as well as excluding parts of the ligand from surrounding solution.
Figure 3: Close-up on alpha-Amanitin with contact residues to the designed binder including hydrogen bonding (orange dashes: hydrogen bonds; blue dashes: strict hydrogen bonds Our part includes the coding sequence for expressing design #15 in E.coli. Additional information and resources are available on our Gitlab https://gitlab.igem.org/2025/software-tools/hamburg, including:
- Alpha-Amanitin binder protein sequence
- Boltzdesign1 output structure
- Re-prediction structures from Alphafold3 and Boltz2
- Data table with all quality metrics and prediction statistics
Our hope is that future iGEM teams can use the information provided here as a starting point for further structural design or directly for recombinant expression. To access the part we refer to the iGEM registry at: https://registry.igem.org/parts/bba-25owwae9
Beyond our approach of helping with death cap intoxication, we see uses in cancer therapy as an amanitin delivery system or as a scaffold for enzyme design.