ENGINEERING
Overview
During the need-scoping process of our integrated human practices, we engaged with a variety of professionals and literature, to isolate 3 areas of focus we could develop a solution for
- PFAS prevention: developing alternate materials to use instead of PFAS
- PFAS contamination: Finding ways to limit PFAS exposure through filtration
- PFAS bioaccumulation: determining a means to remove PFAS from the system
Throughout the process of our protein design, we engaged with the guiding engineering principles of iGEM Design, Build, Test, and Learn in the development of a solution. Our initial ideas for a solution were inspired by nature, specifically, IgA antibodies. IgA antibodies enjoy high specificity and binding strength associated with antibodies, but also have the ability to be secreted from the bloodstream, where they are made, to the mucosa, where they serve as an immune defense.
We wanted to create a protein with the same abilities as an IgA antibody: something with high specificity and ability to be transcytosis from the blood to the mucosa. We first tried to create a humanized antibody with a novel Fab site for PFOA. From our initial literature review, we learned that secretory IgA (sIgA) is a dimer, joined by a J chain peptide. The J chain peptide not only provides a physical scaffold to link the two IgA monomers into a dimer, but is also essential in transcytosis from the bloodstream to the mucosa, which is mediated by the polymeric immunoglobulin receptor (pIgR). pIgR, which is mostly found on mucosa cells (1), recognizes the J chain of sIgA dimers, which then signals associated transcytosis events in the mucosa cell.
Designs
Design 1: A new IgA Antibody
Build: We attempted to build a novel Fab using OptMAVEn-2.0 (2).
Test: When using OptMAVEn-2.0, we ran into an impassable error since we were trying to build the antibody off non-protein residues (the PFOA molecule)
Learn: We learned from further literature review that antibodies don’t typically bind to small molecules. We were not going to be able to use a traditional Fab. From our literature review, we learned that albumin is a natural protein that has the ability to bind to PFOA with a respectable binding affinity in the micromolar range. We then sought to complex the pocket of Albumin that binds to PFOA to a Fc domain region of antibody, since the Fc region of the antibody is where the J chain attaches and how pIgR mediates transcytosis.
Design 2: A PFOA pocket complexed to human Fc
Build: We designed a protein sequence for a pseudo-antibody that had a N→C continuous pocket of albumin that was found to bind to PFOA. The pocket cutoff parameters was simply that the secondary structures which contained the PFOA binding residues (which were identified in the paper which did crystallized the Albumin-PFOA complex) At the C terminus of this pocket we appended the hinge and Fc domain from a human IgA antibody. To build this model, we created the sequence and then predicted its structure with Alpahfold 2.
Test: Before we ordered the gene sequence, we discussed it with our scientific advisors, and realized it would be extremely difficult to express IgA proteins in E.coli due to E.coli’s lack of post-translational protein machinery, the protein’s large size which could stress E. coli, and abundance of disulfide bonds, which are very difficult to form in the reducing environment of the E.coli. cytoplasm.
Learn: We did further literature review. One option we considered was trying to express the chimeric protein in a new, more complex, expression system like yeast– but this would require considerable upfront costs. Fortunately, we found a paper describing an attempt to harness the transcytosis ability of pIgR through the attachment of a small peptide that could be attached to the C-terminus of an IgG antibody. We contacted the authors of the paper and set up a meeting with him. He told us how based on the results, we could get the transcytosis effect we desired with the minimal peptide sequence his lab used, as compared to the Fc domain.
Design 3: A PFOA pocket with a PASR peptide
Build: We removed the hinge and Fc domain from Design 2 and replaced it with the peptide QRNPKLKLIRRHPTLRIPPI. The original moniker for this sequence was QRN, but for our project we have called it PASR (Peptide thAt triggerS pIgR-mediated transcytosis). The PASR peptide has been shown to bind to human and mouse pIgR at comparable affinities in-vitro. The PASR sequence has been shown to bind to pIgR and facilitate transcytosis when complexed to c-terminus of IgG in mice. The peptide has been optimized to express well and not aggregate.
Test: We used the linker sequence used in the original paper, GGGGS, to link the PFOA-binding pocket to the PASR peptide. We created the sequence and appended it to both the N and C termini of the pocket and then folded the sequences with Alpha Fold 2. Learn: Likely due to its small size and abundance of charged residues, alpahfold2 predicted the PASR to have no secondary structure. For this reason, for further designs we decided to append the PASR to the N-terminus, as we did not want it to destabilize the C-terminus alpha helix (which has some residues in the binding pocket.) The N-terminus is further away from the binding pocket than the C-terminus.However, when reviewing the literature once more before ordering, we learned of a critical flaw in our design. Albumin’s 4 pockets for PFOA were not equal binding strength; one pocket had significantly higher binding affinity than the other 3. The pocket our chimeric protein was built off of was the low affinity binder.
Design 4: High affinity PFOA pocket with a PASR peptide
Build: We isolated the correct, high-affinity pocket of Albumin, which luckily enough was continuous from N→C. We could modify our design by simply replacing the old pocket with the new pocket. Like the first design, we defined our pocket to be the minimal number of residues needed to not disturb any secondary structures which the binding amino acids were contained in.
Test: We tested appending the PASR peptide to both the N and C terminus. By this time, Alpha Fold 3 had been released, so we used Alpha Fold 3 to fold designs with the PASR sequence appended to either the N or C terminus with the GGGGS linker.
Learn: In this pocket, both the N-terminus and C-terminus were predicted to be unstable when the PASR peptide was appended. This would be problematic since the N and C termini were both very close to the binding pocket, and an intrinsically disordered region (IDR) near this could cause the binding pocket to be unavailable as the IDR flailed around.
Design 5: Expanded High-affinity PFOA pocket with a PASR peptide
Build: We could either expand the N or C terminus to add more space for the PASR peptide to move around without blocking the binding pocket. We decided to expand the C terminus for multiple reasons:
- The residues further along Albumin in the C terminus direction faced the solvent, while the residues further along Albumin the N terminus direction made even more intra-protein contacts. We reasoned that residues that natively face water would be more stable than residues that make abundant intra-protein contacts that would be absent in our minimal pocket structure.
- Expanding residues further along Albumin in the N terminus direction could increase the chances of unwanted fatty acid binding, since in the Albumin crystal structure we built our protein off of, myristic acid was observed in the FA3 pocket; we did not want to add more residues of the FA3 pocket and increase the likelihood of off-target binding.
Test: We tested multiple lengths of an expanded c-terminus based on the same principle of keeping secondary structures intact. We predicted stability of the proteins by predicting the structure and assessing them in Alpha Fold 3. We predicted the structure of both the binding pocket with the PASR and linker on the N and the newly expanded C terminus.
Learn: The pocket was most stable when we expanded the pocket structure by adding 2 alpha helices on the c-terminus side from the real albumin. Importantly, these residues have not been shown to bind to any other ligands, and resulted in a PASR peptide that was ~110% more further away from the PFOA ligand (~11 Angstrom vs ~23 Angstroms.)
However, an old issue came up in our new design: there were methionines, cysteines and a disulfide bond in the pocket, which as previously mentioned, can cause folding issues when E.coli is used as the expression system.
Design 6: LigandMPNN Redesigned PFOA+PASR chimera protein
Build: We wanted to see if LigandMPNN could redesign the PFOA pocket and increase its binding affinity and also optimize for protein stability by removing cysteine and methionine. We input Design 5 into LigandMPNN with the default parameters, but restricted cysteines and methionines from being in our designed peptide sequences.
Test: Before we ordered the designs, we realized we would not have the wet lab resources needed to properly utilize LigandMPNN, which requires considerable wet-lab based optimization after an initial screening is done
Learn: Although we could not use LigandMPNN to increase binding affinity, we could still use it to redesign the non-binding residues for stability by removing the cysteines and methonines, which as previously mentioned cause protein folding issues in E.coli expression systems. The proof of concept for this approach has already been tested and requires only a fraction of the wet lab testing, which was a workload suitable for our lab.
Final Design (Design 7): LigandMPNN Partial Redesigned PFOA+PASR chimera protein
Build: We put Design 5 through the LigandMPNN algorithm, but fixed all residues within 7 angstroms of the PFOA ligand (except two cysteines that formed a disulfide bridge). We once again restricted cysteine and methionine from being incorporated into the generated protein sequences and took the fixed residues as context to build the rest of the protein.
Test: LigandMPNN is a stochastic algorithm, meaning it does not converge on just one solution, but rather generates many sequences that could work. We folded 120 of these designs with Alpha Fold 3 and selected the top 4 designs with the lowest RMSD, highest pTm and highest plDDT scores to order for wet lab testing.
Final Design Scheme: Residues colored green represent amino acids that were kept fixed during the LigandMPNN procedure.
Design 8: ELISA Sandwich Assay to test PFAS Binding affinity
Build: Our initial plan involved using a sandwich ELISA to assess the binding affinity between our engineered binder, containing the FA4 pocket and PASR sequence, and PFOA/PFOSA. PFOA/PFOSA would be immobilized onto the ELISA plate (blue box) via physical absorption. The binder, which included a his-tag (black) from the plasmid for detection, would then be introduced. We intended to detect the binding event using an anti-his-tag antibody (yellow), followed by an anti-mouse IgG (brown) conjugated to an enzyme (red) that would catalyze a fluorescent reaction upon binding with substrate (purple). This fluorescence would provide a quantitative measurement of the binder's ability to interact with PFOA/PFOSA.
Test: We planned to perform the ELISA in a sandwich format, where PFOA/PFOSA would be immobilized on the plate, and the engineered binder, with its his-tag, would bind to these molecules. The detection mechanism would involve a secondary antibody system to amplify the signal, providing measurable fluorescence when successful binding occurred. A control group using unmodified albumin would establish a baseline for comparison.
Learn: However, through advice from a mentor, we quickly realized that the ELISA assay was not suitable for our purpose. The main issue was the size of the PFOA/PFOS molecules, which are significantly smaller than typical antigens used in ELISAs. For example, insulin (a small protein) is around 6 kDa, while PFOA is roughly 200 Da. ELISA is more appropriate for larger proteins, and the size of perfluoroalkyl compounds like PFOA makes it unlikely they could be effectively immobilized for a sandwich assay. Additionally, Biolayer Interferometry (BLI) faced similar challenges due to the small size of the molecules and the expense and availability of BLI tips. Based on this insight, we abandoned the ELISA and BLI approaches.
Instead, we shifted our focus to Differential Scanning Fluorimetry (DSF), which allows us to assess the thermal stability of the binder in the presence of PFOA. In DSF, we monitor tryptophan fluorescence as the protein is heated, tracking the transition from folded to unfolded states. A higher melting temperature (Tm) indicates greater stability, meaning better binding to PFOA. Our goal is to see a high Tm for PFOA, showing strong interaction with the binder, and a lower Tm for myristic acid, serving as a control to ensure specificity.