The efficiency of defensins to form polymers is an important determinant of its antibacterial property (Rajabi et al. 2012). For example, the HNP1 mutants W26A and MeIle-20 have weakened antibacterial activity (Pazgier，2012). Defensin dimers serve as an important and fundamental building block for more complex polymers. Thus, in the modelling section, we studied the process of dimerization in alpha-defensins, in which HNP-1 was studied, and instigated point mutations to improve the binding efficiency of defensins.

α-defensins typically have a triple-stranded beta-sheet structure based on internal disulfide bonds (Lehrer et al. 2012) and further stabilized by other immutable structures. Apart from these residues, point mutations on alpha-defensins tend to have limited effect on the general conformation. Alpha-defensin dimers are formed when amino acids on the second beta-strand interact with each other. Specifically, carboxyl and nitric groups in the amino acid backbone of the two monomers formed 4 hydrogen bonds during dimerization. In addition, cationicity and hydrophobicity are both key determinants in dimerization rates, with the latter showing more significant impact than the former (Rajabi et al. 2012). Thus, our general goal is to create a stable dimer structure while increasing cationicity and hydrophobicity.

We begin the analysis and design of HNP-1 through analysis of the conformation of the wild type HNP-1 through ClusPro protein docking analysis. HNP-1 dimer models predicted by ClusPro are compared to the structures of dimer molecules in pre-existing research to ensure the accuracy and possibility of utilizing protein engineering.

The ClusPro balanced model provided various clusters for HNP-1 homodimerization, where both electrostatic and hydrophobic interactions are taken into account. The energy scores of each cluster indicate the stability of the dimer configurations, with lower energy values corresponding to more stable dimers. Below are the results of the most relevant clusters, focusing on the center energy and the lowest energy conformation within each cluster. Results of ClusPro Analysis are as follows:

Cluster 0

- Center energy: -966.6 kcal/mol
- Lowest energy: -1083.1 kcal/mol
- This cluster provides one of the most stable configurations, with a significantly low lowest energy score, suggesting strong interactions between the HNP-1 monomers.

Cluster 1

- Center energy: -1078.2 kcal/mol
- Lowest energy: -1078.2 kcal/mol
- The energy scores for Cluster 1 indicate a highly stable dimer with little internal variation, further supporting the stability of its configuration.

Cluster 2

- Center energy: -873.4 kcal/mol
- Lowest energy: -1160.2 kcal/mol
- This cluster shows the lowest energy conformation across all simulations, making it a promising candidate for further exploration of stable dimer interactions, though the center energy reflects variability within the cluster.

Clusters 0 and 1 also demonstrate highly stable dimer formations, with Cluster 1 being particularly consistent due to the close match between its center and lowest energy scores. The lowest energy conformation in Cluster 2 (-1160.2 kcal/mol) is the most stable among all clusters, indicating a strong affinity between the two HNP-1 monomers in this configuration. In fact, Cluster 2 is consistent with reported crystal structure of HNP-1 dimers. In addition, the high levels of variability (large difference between the center and lowest energy) illustrates high possibility and opportunity of refining interactions between the dimers through point mutations that would create more stable configurations while also suggesting higher flexibility of binding conformations that would reduce the risk of functionality loss due to mutated active sites. Therefore, we confirmed the possibility of using of modeling to optimize and stabilize dimer interactions.

Our models can mainly be categorized in active site amino acid mutations and other amino acid mutations. We sought to stabilize the active site by introducing additional interactions between the two monomers. Upon considerations, we decided attempt creating pi-pi, pi-cation, and hydrogen bonding interactions between amino acid side chains. Aromatic rings composes of conjugate pi-bonds, which causes the pi-electrons to be delocalized and shared across multiple molecules. A pi-pi interaction is when the delocalized electrons are shared between two aromatic rings when placed adjacently, while pi-cation interaction is characterized by an attraction between a positively group and the delocalized pi-electrons. It should be noted that mutagenesis in our models might not accurately reflect the actual configuration of side chain of the modified HNP-1 that are produced in experimentation. Therefore, we would prioritize pi-pi interactions (increase hydrophobicity) and pi-cation (increase hydrophobicity and cationicity) over hydrogen bonding under the same predicted affinity. The tetramer of HNP-1 was obtained from the Protein Data Bank (PDB id: 3GNY), and the dimer molecule was extracted. The amino acids of the active site are T-18, C-19, and I-20 (Amino acid - index), and as C-19 is immutable (as discussed later), we will focus on mutating T-18 and I-20.

Apart from active site amino acids, other amino acids were also mutated to increase hydrophobicity. It is important to note that there are several immutable that constructs the basic structure of alpha-defensin, namely the cytosine residues (C-2, C-4, C-9, C-19, C-29, C-30) that constitute disulfide bonds, arginine and glutamic acid residues (R-5, E-13) that constitute an essential salt bridge, and an invariant glycine (G-17). Of the remaining amino acids, we selected all non-positive hydrophilic amino acids and mutated them to alanine to increase hydrophobicity without drastically changing the general conformation.

Active site mutations were achieved through PyMOL mutagenesis. Amino acid mutation orientations were selected in attempt to create pi-pi, pi-cation, or hydrogen bonding. Note that the conformations of the monomers selected should be identical as the monomers should be structurally identical. For alanine substitution, specific amino acids on HNP-1 are directly changed by rewriting the sequence and the protein conformation is determined through alpha-fold.

The same monomers were inputed into the Hdock server as both the receptor and ligand. To ensure that the returned results form dimer structures as current research indicates, the receptor and ligand binding site residues is entered as 18-20:A, with residue distance restraint entered as 18:A 20:A 3.4, 19:A 19:A 3.4, 20:A 18:A 3.4 to ensure that side chains interacts with the backbones at a distance typical for hydrogen bond. The Hdock result with the correct dimer structure were analyzed and the confidence score is used as standards for the binding affinity. The original model’s score is 0.7116, so a mutation with a higher score is considered successful.

Pi-pi interactions

Pi-pi interactions between R groups of amino acids can be achieved through mutations to change T-18 and I-20 into aromatic amino acids tryptophan (W) and phenylalanine (F). The 18th amino acid on one strand is directly adjacent to the 20th amino acid on the other strand. Thus, amino acids can interact across (18 to 18 or 20 to 20) or adjacently (18 to 20).

Through observations of mutagenesis result, we concluded that 18 to 18 and 18 to 20 mutations either have great clashes or have limited interaction domains. Thus, we shift our attention to mutations at the 20th amino acid. Pi-pi interactions are observed in both models with little clashing in the configurations. Hdock confidence scores are 0.7657 and 0.7807 for HNP-1_I20F and HNP-1_I20W, respectively. The higher confidence score given for tryptophan mutation is likely resulted from greater hydrophobicity due to a larger aromatic ring.

Pi-cation interactions

Pi-cation interactions between R groups can only be achieved through interactions between adjacent base pairs (18 to 20). The mutation HNP-1_AWK (T18W_I20K) showed greatest similarity with previously known pi-cation interaction between acetylcholine and tryptophan.

Hydrogen bonding

Both the 18th and 20th amino acid were mutated to Asparagine to create hydrogen bonds between the side chains, which was demonstrated in PyMOL after mutagenesis. Hdock results (confidence score = 0.7365), however, showed that the side chain of 18N created hydrogen bonds with the carboxyl group of the amino acid backbone on the other monomer while the side chain of 20N did not have any interactions with nearby residues. Therefore, to increase hydrophobicity meanwhile inducing pi-pi interactions as well, the the 20th amino acid was mutated to F (20W causes clashing with 18N). The final model HNP-1_ANF (T18N_I20F) obtained a confidence score of 0.7947.

Other amino acid modifications

After the alanine substitution of amino acids, the alpha-fold predicted model shows high similarity with the original HNP-1 when aligned in PyMOL (rmsd = 0.482). Hdock result of the final structure HNP-1_Ala gives a confidence score of 0.8058.

Based on the aforementioned modeling predictions, we decided to synthesize 4 HNP-1 mutants based on the 4 major mutations (HNP-1Ala: Alanine substitution; HNP1AWW: Pi-pi; HNP1AWK: Pi-cation) using our CBM3-SUMO-Defensins expression module in Design3. The sequences of all HNP1 mutants are shown in fig.4.

Expression of HNP1 Mutations

The DNA fragments of the HNP1 mutants were obtained through SOE-PCR, as described in the [protocol](link to Experiment). These fragments were ligated with the digested pLINKS2400 expression vector via golden gate assembly. Escherichia coli DH5α competent cells were transformed and subsequently grown in LB medium containing kanamycin. The plasmids were validated through colony PCR and DNA sequencing to ensure the correct genetic sequence. The confirmed plasmids were named pLINKS2439, pLINKS2440, pLINKS2441, and pLINKS2442, corresponding to the constructs CBM3-SUMO-HNP1Ala, CBM3-SUMO-HNP1ANF, CBM3-SUMO-HNP1AWW, and CBM3-SUMO-HNP1AWK, respectively. These plasmids were then transferred into E. coli SHuffle T7 competent cells for protein expression.

To optimize expression, E. coli SHuffle T7 cells containing plasmids pLINKS2439 to pLINKS2442 were first cultured in fresh LB medium overnight at 37°C. The following day, the cultures were subcultured into 4 mL LB medium with kanamycin and incubated on a rotary shaker at 30°C for 3.5 hours, until OD600 reached 0.6 to 0.8. Cells were harvested by centrifugation, resuspended in 20 mM Tris-HCl buffer, and lysed. The expression of HNP1Ala, HNP1AWW, and HNP1AWK was successfully confirmed via SDS-PAGE analysis. We then scaled the expression up to 400 mL flask cultures and purified the defensin mutants using Ni-NTA affinity chromatography from the cell lysates.

Antibacterial activity of HNP1 mutants

We successfully purified the fusion proteins except CBM3-SUMO-HNP1ANF and used the same program (a link leads to Engineering Success) for sumo digestion, and then measured the 12-hour growth curve and 8-hour MIC of Staphylococcus aureus. The results are shown in Figure 6.

The results indicated that:
- The MIC99 for HNP1 was 2.37 μM,
- For HNP1AWW, it was 3.53 μM.
- The MICs for HNP1Ala and HNP1AWK were higher than the selected concentrations, suggesting that the mutations significantly reduced the antibacterial activity against Staphylococcus aureus.

The results are generally unsuccessful, with our mutations generally resulting in reduced antibacterial activity. We hypothesize several factors in our modeling process that might contribute to this failure. Firstly, defensin mechanism extends beyond dimerization, including binding to Lipid II (key metabolic component of bacterial cell wall formulation) and other LF factors that reduces bacterial growth. Although most mutations would not significantly alter the structure of defensin, it is necessary to note that the immutable glycine which plays a significant role in LF factor binding is located at the 17th amino acid and the mutations at the 18th and 20th amino acid might therefore interfere with antibacterial effects. In addition, the structures of mutants visualized in this project might illustrate deviations from the true conformation, especially considering the fact that mutated conformations were chosen based on mutagenesis without affirmation through structure prediction and refolding (e.g., Alphafold). This might cause the structures to actually not be further stabilized by pi-pi, pi-cation, or hydrogen bonding, and for the example of hydrogen bonding mutations, dimerization efficiency might even be reduced due to reduced hydrophobicity. Lastly, dimer structures created in this project might experience clashing between R groups of mutated amino acids due to its general greater sizes.

Certain solutions are proposed for these problems. For example, docking analysis between defensin and Lipid II/LF factors could be performed to mitigate side effects of the current mutations and even improve antibacterial activity. Furthermore, the inclusion of other mutations with smaller R groups bases (e.g., Phenylalanine) might be able to reduce interference with LipidII/LF factor binding and clashing, which can be further verified using protein folding algorithms. Deeper research on the field of protein engineering also suggest alternative methods of optimization, including the use of AI to generate “smarter” models and mutations.

Nevertheless, our project provides a working system and procedure for the use of protein engineering in optimizing protein activities and characteristics. Though results were not ideal, the presence of antibacterial activity of our mutants (HNP1Ala is even be better at 0.59 μM) demonstrates the possibility of using mutant protein substitutions to achieve specific goals. Our project demonstrated that mutations of non-immutable amino acids could change protein characteristics without deteriorating its original function and conformation, which could powerful in artificially adding multiple favorable properties to proteins. Applications could include but is definitely not limited to creating better performing proteins at artificial environments which deviates from proteins’ natural environment, attributing new functions and even metabolic activities, and understanding specific metabolic or enzymatic mechanism.