Model

Overview

As part of the drylab we wanted to further investigate the adhesion quality of chitin to chitinase-92 and the synthetic peptide 1. To do this, we created structure models of the protein's binding sites with Alphafold2 then created a reduced model of chitin to use as a ligand during docking simulations with Autodock.

Structure Model

To genetically alter Bacillus subtilis or another organism to adhere better to Candida albicans, we want to express the binding domain of chitinase-92 on the organism's surface. Chitinase-92 is a known enzyme that has a double chitin binding domain ChBD__CICII corresponding to the protein's sequence domain (Asn769 to Asn865 of Chi92) ¹ and therefor if expressed on the outer membrane of B. subtilis would establish an adhesion quality of the bacterium. In the table below, the amino acid sequence of the double chitin binding domain from chitinase-92 and a synthetic peptide (SP1) are given. SP1 is an amino acid sequence that was analyzed in previous research to compare the adhesion quality to chitin and showed promising binding qualities², which is why we chose to consider it as well in our experiments.

Since neither for chitinase-92 nor SP1 experimental data of their structure exists, we used the ColabFold ³ template to predict the structure of the binding sites given by the amino acid sequences with Alphafold2 ⁴. For the chitinase-92 prediction we gave the model templates of the chitin binding family 5/12 with a resolution lower than 2.5Å from the RSCB Protein Data Bank⁵, since for those experimental data does exist and the binding domain is similar to the one of chitinase-92. This is particularly helpful for the model to estimate its confidence in the prediction.

In the figures below, the results of the structure predictions are shown. The cartoon models of the predicted binding domains, including their secondary structure elements and color-coded pLDDT scores, were visualized in ChimeraX⁶ using the PDB output from ColabFold. The additional plots correspond to the standard ColabFold outputs and help interpret the reliability of the predictions. The first panel shows the highest-ranked predicted structure, followed by the per-residue pLDDT scores across all five models, which indicate the model’s confidence in local atomic positions. The third plot displays the Predicted Aligned Error (PAE) for the best-ranked model, visualizing the expected positional uncertainty between residue pairs and highlighting regions with potential flexibility. Finally, the sequence coverage plot summarizes which sequence regions were aligned to structural templates during prediction, providing context for the model’s confidence in different parts of the protein.

1 / 4

2 / 4

3 / 4

4 / 4

❮ ❯

The predicted structure of the chitinase-92 binding domain shows lower confidence in the thin, flexible regions connecting the denser secondary structure elements, as expected for highly mobile loops. In the denser regions, pLDDT scores are generally above 70, indicating reliable structural predictions. The per-residue pLDDT plots are consistent across all five predicted models, highlighting the same low-confidence regions at the connecting elements and the C-terminus of the binding domain. This symmetry between the two dense regions is also reflected in the PAE plot. Although the sequence alignment plot indicates limited template coverage for the input sequence, the overall confidence scores remain reasonable, likely because templates for the entire chitin-binding family 5/12 proteins were used, rather than only the binding domain, which reduces apparent coverage in the alignment.

1 / 4

2 / 4

3 / 4

4 / 4

❮ ❯

The structure prediction of the SP1 binding domain shows overall low confidence, which is expected given the lack of suitable structural templates. The sequence coverage plot confirms that no homologous templates were identified, as the model only references the SP1 sequence itself. Consequently, the PAE plot indicates reliable predictions only for short local interactions between neighboring residues, while global structural confidence remains low. Since SP1 is a synthetic random peptide designed merely to match the length and amino acid composition of the chitinase-92 double binding domain, this uncertainty is not unexpected. Further structural modeling would therefore require experimental validation, and the subsequent docking analysis should be interpreted with caution due to the limited confidence in the predicted conformation.

Ligand Model

The target structure for adhesion on the surface of C. albicans is chitin, a long-chain polymer composed of N-acetylglucosamine (NAG) monomers linked by β-1,4-glycosidic bonds ^{7, 8}. Because modeling an entire polymer is computationally demanding and unnecessary for studying local interactions, we instead used shorter oligomers—chitobiose⁹ and chitotriose¹⁰—as ligand models. These naturally occurring fragments retain the key functional groups of chitin, such as hydroxyl and acetylated amine groups, in the correct spatial orientation. Moreover, short oligomers like these have been experimentally observed to bind to chitinases and chitin-binding domains in X-ray structures and biochemical assays, such as the family 5/12 chitinase with PDB ID 4HME ¹¹.

Chitobiose Chitotriose

Monomer Dimer Trimer

Since chitobiose and chitotriose contain open-ring configurations that are less frequently present in natural chitin structures, we modeled dimers and trimers of N-acetylglucosamine (NAG)⁸ linked by β-1,4-glycosidic bonds to better represent the predominant closed-ring structure of a chitin surface. The NAG monomer structure was obtained from PubChem, and the dimeric and trimeric ligands were constructed in Avogadro 2^12,13. As with chitobiose and chitotriose, we used Avogadro’s geometry optimization tool to minimize the molecular energy and adjust bond lengths to physically realistic values, ensuring that the final ligand geometries accurately reflected biologically relevant conformations. All the structure formulas were created in MolView¹⁴.

Docking Simulations

To evaluate the binding of SP1 and the chitinase-92 binding domain to chitin-related ligands, molecular docking simulations were performed using AutoDock 4.2.6 ¹⁵. The protein binding sites were treated as rigid macromolecules, while only the ligands were allowed to rotate around their flexible bonds. To reduce computational cost without compromising accuracy, all amide bonds were fixed as non-rotatable, reflecting their inherent structural rigidity. Docking simulations were carried out using AutoDock’s genetic algorithm with default parameters, except for adjustments to the population size, number of generations, and number of runs as listed in the table below. For ligands with a greater number of rotatable bonds, the parameters were scaled to ensure sufficient sampling and statistical reliability. The docking grid was defined to fully encompass the binding site of each macromolecule, with a grid spacing of 0.375 Å in the AutoDock coordinate system.

Results

The docking simulations provided estimates of the free energy of binding for each ligand–protein complex. In AutoDock, the free energy of binding is calculated as the sum of the final intermolecular energy, the final total internal energy, and the torsional free energy, minus the energy of the unbound system. The table below lists the minimum binding energy values obtained from the best run of each docking simulation.

Overall, both chitinase-92 and SP1 displayed lower free binding energies for the closed-ring ligand structures, indicating stronger binding and higher adhesion potential. SP1 exhibited significantly lower binding energy with chitobiose compared to chitinase-92, supporting previously observed strong experimental adhesion results ². In contrast, the SP1–chitotriose docking produced a relatively poor result, likely due to limitations in the rigid macromolecule model and the chosen docking parameters, which may have prevented the ligand from finding an optimal binding orientation within 50 runs. Visualization of this dock showed that chitotriose did not occupy a pocket as it did with chitobiose. For the closed-ring NAG dimer and trimer, both chitinase-92 and SP1 demonstrated promising binding energies of a similar magnitude. None of the ligands bound to the planar β-sheet region of SP1 that was hypothesized to be a favorable carbohydrate-binding site; in our rigid model frame, these sheets were buried within the protein and inaccessible. Notably, other binding pockets with favorable binding energies were identified. In chitinase-92, the preferred binding site was located at the interface between the two dense structural regions of the binding domain.

Below you can find images of the modeled dock with the binding pocket and the interaction analysis upon hovering. All these images and analysis were created with PyMol¹⁶.

Docking result chitinase-92 and trimer

Docking result SP1 and trimer

Docking result SP1 and chitobiose

None of the ligands bound to the planar β-sheet region of SP1 that was hypothesized to favor carbohydrate interactions; in the rigid model frame, these sheets were largely buried within the protein and only partially accessible. However, alternative binding pockets with favorable binding energies were identified. For chitinase-92, the preferred binding site was located at the interface between the two dense structural regions of the binding domain. The depicted docking pose with the lowest free binding energies showed a polar interaction with an aromatic residue, a feature previously proposed to enhance carbohydrate binding, thereby supporting this hypothesis.

References

1. Ming Chung Chang, Pe Lin Lai, Mei Li Wu, Biochemical characterization and site-directed mutational analysis of the double chitin-binding domain from chitinase 92 of Aeromonas hydrophila JP101 , FEMS Microbiology Letters, Volume 232, Issue 1, 2004, Pages 61-66, ISSN 0378-1097
2. Chamas A, Svensson CM, Maneira C, Sporniak M, Figge MT, Lackner G. Engineering Adhesion of the Probiotic Strain Escherichia coli Nissle to the Fungal Pathogen Candida albicans. ACS Synth Biol. 2024 Dec 20;13(12):4027-39.
3. Mirdita, M., Schütze, K., Moriwaki, Y., et al. (2022). ColabFold: making protein folding accessible to all. Nature Methods, 19, 679-682.
4. Jumper, J., Evans, R., Pritzel, A., et al. (2021). Highly accurate protein structure prediction with AlphaFold. Nature, 596, 583-589.
5. H.M. Berman, J. Westbrook, Z. Feng, G. Gilliland, T.N. Bhat, H. Weissig, I.N. Shindyalov, P.E. Bourne, The Protein Data Bank (2000) Nucleic Acids Research 28: 235-242 https://doi.org/10.1093/nar/28.1.235.
6. UCSF ChimeraX: Tools for structure building and analysis. Meng EC, Goddard TD, Pettersen EF, Couch GS, Pearson ZJ, Morris JH, Ferrin TE. Protein Sci. 2023 Nov;32(11):e4792.
7. PubChem [Internet]. Bethesda (MD): National Library of Medicine (US), National Center for Biotechnology Information; 2004-. PubChem Compound Summary for Chitin; [cited 2025 Oct. 7]. Available from: https://pubchem.ncbi.nlm.nih.gov/compound/Chitin
8. PubChem [Internet]. Bethesda (MD): National Library of Medicine (US), National Center for Biotechnology Information; 2004-. PubChem Compound Summary for CID 24139, N-ACETYL-beta-D-GLUCOSAMINE; [cited 2025 Oct. 7]. Available from: https://pubchem.ncbi.nlm.nih.gov/compound/N-ACETYL-beta-D-GLUCOSAMINE
9. PubChem [Internet]. Bethesda (MD): National Library of Medicine (US), National Center for Biotechnology Information; 2004-. PubChem Compound Summary for CID 122158, Chitobiose; [cited 2025 Oct. 7]. Available from: https://pubchem.ncbi.nlm.nih.gov/compound/Chitobiose
10. PubChem [Internet]. Bethesda (MD): National Library of Medicine (US), National Center for Biotechnology Information; 2004-. PubChem Compound Summary for CID 121978, Chitotriose; [cited 2025 Oct. 7]. Available from: https://pubchem.ncbi.nlm.nih.gov/compound/Chitotriose
11. Malecki, P.H., Raczynska, J.E., Vorgias, C.E., Rypniewski, W. Structure of a complete four-domain chitinase from Moritella marina, a marine psychrophilic bacterium , (2013) Acta Crystallogr D Biol Crystallogr 69: 821-829
12. Avogadro 2 [Internet]. [cited 2025 Oct 7]. Available from: https://www.openchemistry.org/projects/avogadro2/
13. Hanwell, M., & Hutchison, G. Avogadro [Computer software]. https://github.com/openchemistry/avogadrolibs
14. MolView: A web-based molecular visualization tool. [Internet]. [cited 2025 Oct 7]. Available from: https://app.molview.com/
15. Morris, G. M., Huey, R., Lindstrom, W., Sanner, M. F., Belew, R. K., Goodsell, D. S. and Olson, A. J. (2009) Autodock4 and AutoDockTools4: automated docking with selective receptor flexiblity. J. Computational Chemistry 2009, 16: 2785-91.
16. Schrödinger, LLC. (2015). The PyMOL Molecular Graphics System, Version 1.8. Retrieved from Schrödinger, LLC.