Results

The Results section brings together all the work we performed --- from computational protein modeling to experimental validation --- and shows how each part of the project fits into the larger iGEM cycle of design, build, test, and learn. Our overall goal was to explore whether we could engineer MHC Class I molecules to enhance binding affinity for tumor-associated peptides. This required not only designing mutations in silico, but also cloning, expressing, purifying, and testing recombinant proteins in the lab.

We started with in silico modeling, where we mapped the peptide-binding groove of HLA-A*02:01 and identified key residues responsible for peptide stabilization. Using alanine scanning and docking simulations, we designed two mutant variants that were predicted to alter peptide binding. This computational phase established a rational basis for our experiments and reduced the guesswork in deciding which mutations to test.

Next, we moved into the wet lab phase, where we produced wild-type HLA-A*02:01, Mutant 1, Mutant 2, and β2-microglobulin in E. coli. At first, protein expression was limited by low yields, but by applying colony selection (screening and choosing the best-expressing bacterial colonies), we significantly improved production. Although the proteins localized mostly to inclusion bodies, we successfully purified and solubilized them using urea-based denaturation. This confirmed that bacterial systems can produce engineered human proteins at scale, provided that proper refolding strategies are applied.

The purified proteins were then verified through SDS-PAGE and quantified using BCA assays, which showed that we achieved milligram-per-milliliter levels of recombinant proteins --- enough for downstream testing. Finally, we tested functionality with a fluorescence-based peptide binding assay. This experiment confirmed that wild-type HLA-A*02:01 bound strongly to the pp65 peptide, while Mutant 1 and Mutant 2 both showed reduced binding affinity. While these results did not enhance antigen presentation as originally intended, they proved that structural modifications of MHC molecules can tune binding affinity.

Taken together, our results demonstrate the power of combining computational design with experimental testing. The project highlighted both the potential and the challenges of rational protein engineering: although our specific mutations reduced binding, the findings confirm that binding properties are not fixed, but tunable through deliberate design choices. This insight is crucial for future synthetic biology work aimed at fine-tuning immune responses.

In Silico Results: Structural Analysis of HLA-A*02:01

To begin, we analyzed the structure of HLA-A*02:01, a common allele of MHC Class I. Using the high-resolution crystal structure PDB ID: 6Q3K, we examined the complex of HLA-A*02:01 bound to β2-microglobulin (B2M) and the viral peptide pp65 (sequence: NLVPMVATV).

The α chain of HLA-A*02:01 is divided into three domains. The α1 and α2 domains form the peptide-binding groove, a slot-like pocket that secures short peptides. The α3 domain stabilizes the structure, while B2M ensures proper folding and surface presentation. Inside the groove, we identified 23 residues interacting directly with the peptide, including conserved aromatic residues like Tyr7 and Tyr116 that stabilize hydrophobic anchors, and charged residues like Glu63 and Lys66 that provide electrostatic stabilization.

Figure 1. 3D ribbon illustration of the HLA-A02:01 peptide-binding groove with bound peptide. Key contact residues highlighted.*

This structural map guided our mutagenesis strategy, helping us design Mutant 1 and Mutant 2 by modifying residues predicted to influence peptide stability.

Alanine Scanning and Mutagenesis Predictions

Next, we performed in silico alanine scanning to systematically test the contribution of each residue to peptide binding. Residues that showed significant predicted loss of binding when mutated to alanine were considered essential. In contrast, residues with smaller predicted contributions were considered candidates for modification. This analysis gave us a short list of residues where substitutions could potentially improve peptide affinity without disrupting overall structure.

Docking Simulations with DiffDock

We then employed DiffDock, an AI-based molecular docking program, to simulate peptide--MHC interactions. Using 6Q3K as the receptor and the pp65 peptide as the ligand, we compared the wild-type binding mode with our proposed mutants. DiffDock produced predicted binding poses and affinity scores, allowing us to visualize how mutations might shift the peptide's position or strengthen stabilizing contacts.

Our results suggested that certain mutations in the groove could increase the predicted binding affinity for tumor-associated peptides such as NY-ESO-1 or WT1, which share anchor residue preferences with pp65. These computational predictions served as the basis for designing our experimental constructs.

Protein Expression in E. coli

We next expressed wild-type HLA-A*02:01, Mutant 1, Mutant 2, and B2M in E. coli. Expression was induced with IPTG at mid-log phase growth. SDS-PAGE analysis confirmed robust production of all proteins.

Figure 2. Expression of recombinant proteins in E. coli. SDS-PAGE gel showing lanes for induced lysates, soluble fraction, and inclusion body fraction for HLA-A02:01 and B2M.

Most proteins were present in the inclusion body fraction, a common outcome when expressing eukaryotic proteins in bacteria. However, we noted that not all colonies expressed protein at equal levels. By performing colony selection---screening multiple colonies for expression efficiency and choosing the highest-producing clones---we significantly improved yields. This step reflects the iterative "design--build--test--learn" approach emphasized in iGEM: even small optimizations, like colony choice, can greatly influence experimental success.

Protein Purification and Solubilization

To purify the proteins, inclusion body pellets were isolated by centrifugation and washed multiple times with detergent buffer to remove membrane contaminants. The washed pellets were then solubilized in 8 M urea, which unfolded the proteins completely. Overnight incubation ensured full solubilization, and centrifugation removed any remaining debris.

This approach yielded large amounts of purified, denatured HLA-A*02:01 and B2M. Although the proteins were not yet in a functional, refolded state, this step was critical for obtaining sufficient quantities for downstream refolding and binding studies.

Figure 3. Workflow of protein purification.

Confirmation of Purified Proteins

The purified fractions were analyzed by SDS-PAGE to verify protein recovery. Distinct bands corresponding to the expected molecular weights confirmed the presence of HLA-A*02:01 (25--308 a.a.) and B2M (21--119 a.a.) . Compared to the crude lysates, these purified samples showed much cleaner backgrounds, indicating successful removal of bacterial proteins.

This step demonstrated that, even though the proteins were expressed in inclusion bodies, our purification workflow was effective in producing high-purity samples suitable for biochemical assays.

Figure 4. SDS-PAGE of purified proteins. Gel image with lanes showing purified HLA-A02:01 and B2M alongside molecular weight markers.

Protein Quantification with BCA Assay

We quantified protein concentrations using a BCA assay, which measures the reduction of Cu²⁺ ions by peptide bonds, leading to a color change that can be measured spectrophotometrically. A standard curve was generated using BSA at known concentrations, with the following regression formula:

y = 0.0014x -- 0.0083,

where y is the absorbance and x is the protein concentration .

Figure 5. Graph showing standard curve and measured concentrations for HLA-A02:01 and B2M.

For HLA-A*02:01, absorbance values across dilutions (1/100--1/400) corresponded to an average concentration of 19.3 mg/mL. For B2M, the average concentration was 16.7 mg/mL . These concentrations are relatively high, confirming that our purification yielded sufficient protein for binding assays.

These high concentrations provided ample material for binding assays, showing that our bacterial system was not only efficient but also scalable.

Fluorescence-Based Binding Assay

The final stage of our experiments aimed to confirm whether our purified proteins were capable of binding peptide antigens, and to compare binding strength between wild-type and mutant constructs. To achieve this, we used a fluorescence detection method that allowed us to quantify peptide binding in vitro.

We immobilized our recombinant proteins onto nickel-coated plates. This method takes advantage of the His-tag engineered onto our constructs: the histidine residues have strong affinity for nickel ions, enabling stable and specific attachment of the protein to the plate surface. This ensured that the orientation of the proteins was consistent across wells and that the peptide-binding grooves were accessible for testing.

This assay design provided a direct readout of binding strength: higher fluorescence intensity indicated stronger binding affinity, while lower intensity reflected weaker interactions.

Finally, we tested peptide-binding functionality using a fluorescence detection assay. The wild-type HLA-A*02:01 + B2M complex bound the pp65 peptide with strong fluorescence, confirming correct folding and functional binding.

In contrast, Mutant 1 and Mutant 2 both showed lower fluorescence signals compared to wild type, indicating reduced binding affinity. This was a key finding: while we did not achieve enhanced binding strength as intended, we demonstrated that structural modifications can indeed alter binding.

This result is significant because it shows that the binding affinity of MHC molecules is tunable through 3D structural changes. Even though our mutations weakened binding, the very fact that binding was altered validates our engineering approach.

Figure 7. Fluorescence detection of peptide binding. Bar graph comparing fluorescence intensity of wild-type HLA-A*02:01, Mutant 1, and Mutant 2 complexes with peptide.

Conclusion

Our combined computational and experimental work provided several key insights:

In silico modeling successfully identified peptide-binding residues and guided the design of two mutant constructs.
Colony selection improved expression, highlighting how even practical optimizations contribute to engineering success.
Expression and purification were robust, producing large amounts of recombinant wild-type and mutant proteins despite inclusion body formation.
BCA quantification confirmed high yields, ensuring material availability for downstream studies.
Functional binding was demonstrated: wild-type HLA-A*02:01 bound pp65 peptide strongly, while Mutant 1 and Mutant 2 showed reduced affinity.

Although we did not achieve the intended enhancement of binding strength, our results clearly demonstrate that the binding properties of MHC molecules can be modulated by engineering their 3D structure. This finding is important for future synthetic biology applications, where precise control of antigen presentation could be used to design novel cancer immunotherapies.

Our project therefore stands as a proof-of-concept: computational predictions combined with iterative wet lab optimization can not only reproduce native function but also deliberately alter it. The next challenge is to refine our approach to identify mutations that enhance rather than weaken peptide binding.