Team:St Andrews/Design
Our modelling had three key facets and our part design reflects this, with our part design being guided by Rational Design, our Bioinformatics search, and our RGN protein structural predictor.
Rational Design of Antibody Domains
Our first intent was to work on the invariable components of the antibody, to maximise modularity of our part to other antibody systems. This gave us a choice of using either the CH1, CH2, or CH3 domains for our project. Our supervisor Dr U Schwarz-Linek recommended we use CH3 due to their ease of expression in bacteria – in mind of our Escherichia coli chassis. In his previous experience he’s struggled to express soluble CH1 or CH2 domains in bacteria, and thus with the limited time on our project we ruled out the possibility of re-folding protocols, and chose to only work on CH3 domains.
.The process of rationally designing mutants capable of forming isopeptide bonds followed that of Kwon et al. (2017)1 with the use of structural alignment to guide mutant design. Using the DALI server2, pairwise structural alignments of our wild-type protein and the set of isopeptide domains (IPDs) shown in table 1 were used to identify the IPDs with the greatest structural similarity to our wild-type protein, assessed using the DALI Z-score. IPDs with a Z-score of >5.0 were selected, and structurally super-positioned onto the wild type protein using CCP4MG’s Superpose3 tool, as illustrated in figure 1. Residues in the wild-type protein that matched those forming the isopeptide bond in the Superposed IPD were identified as sites for mutation, the mutation at that site corresponding to the matched bonding residue i.e. if Val 445 in the wild type protein matches Lys 108 forming the isopeptide bond in the IPD, we mutate the 445 position in the wild-type to a Lysine.
Figure 1: A Global structural superposition of Human IgG4 CH3 Domain (PDB ID: 4B53), light blue with Streptococcus gordonii surface protein Sspb C-terminal domain (PDB ID: 2WOY), light green. These alignments were used to compare locations of isopeptide bonds in the bacterial protein to residues on the CH3 domain and identify sites for mutation. B identification of homologous residues in the antibody (blue) with those forming the isopeptide bond in the isopeptide domain (green). Figures made in PyMol.
| Isopeptide Domain (IPD) | Organism | PDB ID |
|---|---|---|
| B repeat units of Collagen-binding protein Cna | Staphylococcus aureus | 1D2P |
| ACE19, the collagen binding subdomain of surface protein ACE | Enterococcus faecalis | 2OKM, 2Z1P |
| SspB C-terminal domain | Streptococcus gordonii | 2WZA, 2WOY |
| Fibronectin binding protein Fbab-B | Streptococcus pyogenes | 2X5P |
| Pilus-presented adhesin, Spy0125 (Cpa), P1 form | Streptococcus pyogenes | 2XI9, 2XIC, 2XID |
| Major pilus backbone protein | Streptococcus agalactiae | 2XTL |
| RrgB Pilus protein | Streptococcus pneumoniae | 2Y1V |
| Major Pilin Spy0128 | Streptococcus pyogenes | 3B2M |
| Major pilin SpaA | Corynebacterium diphtheriae | 3HR6, 3HTL |
| Major pilin subunit BcpA | Bacillus cereus | 3KPT, 3RKP |
| Major Pilin GBS80 C-terminal Fragment | Streptococcus agalactiae | 3PF2 |
| Minor pilin GBS52 | Streptococcus agalactiae | 3PHS |
| Fimbrial adhesin FimA | Actinomyces naeslundii | 3QDH |
| Fimbrial protein FimP | Actinomyces oris | 3UXF |
| Collagen-binding domain of surface protein RspB | Erysipelothrix rhusiopathiae | 3V10 |
| Major Pilin SpaD | Corynebacterium diphtheriae | 4HSS, 4HSQ |
| Surface Protein SPB1 | Streptococcus agalactiae | 4UZG |
| Shaft pilin SpaA | Lactobacillus rhamnosus | 5F44, 5FIE |
| Major pilin backbone protein T-antigen, T13 | Streptococcus pyogenes | 6BBT, 6BBW |
| Collagen Adhesion TIE protein | Bacillus anthracis | 6FWV |
Table 1: Isopeptide domains used for structural comparison to antibody CH3 domains. The protein function and name is given, along with the species of origin. PDB (Protein Data Base) IDs are given for crystal structures used for comparison with CH3 domain structures.
| CH3 Domain (PDB ID) | PDB ID of IPDs with >5.0 Z-score |
|---|---|
| Human IgG4 (4B53) | 2WOY, 4HSS, 3KPT |
| Mouse Mak33 IgG (1CQK) | 4HSS, 3KPT, 2XTL, 2OKM, 3UXF |
| Human IgE (5MOL) | 4HSS, 4UZG, 3KPT, 2XTL, 3UXF, 2WZA |
Table 2: CH3 Domains chosen for editing and the Isopeptide Domains (IPDs) that return a Z-score of >5.0 when aligned with that CH3 domain using a DALI pairwise search2.
Similarly, we also matched residues in the wild type to any aromatic residues (Tryptophan, Tyrosine, Phenylalanine) adjacent to the bond in the IPD. We see close association in most (but not all) isopeptide bonds between the bond and an aromatic group, the ring structure being typically planar with respect to the bond. Additionally, Kwon and colleagues introduced a previously absent Phenylalanine into their successful mutant, suggesting to us that this aromatic feature is significant to bond formation – perhaps due to a role in excluding water from the reaction, something shown to be significant to bond formation4,5. As such, we included an adjacent aromatic in all mutants, with the final mutant motif being a Lys, Asn, Asp, and a Tyr.
Asparagine was selected as a bonding residue in all mutants due to the slightly preferential properties of Ammonia as a leaving group4, and that all matched IPDs contained Asn as a bonding residue. Similarly, Tyrosine was the aromatic group included in all mutants due to all bar one of the matched IPDs containing a Tyrosine as the adjacent aromatic to the bond.
CH3 Domains from the Human IgE, and the Mouse (Mus musculus) and Human IgG antibodies were obtained from the Protein Data Bank (PDB), with the corresponding PDB codes for the structures used shown in table 2. Upon inspection of the structures we identified a conserved Leucine residue (Leu 441 in Human IgG, Leu 102 in Mouse IgG) which could sterically prevent our proposed Lysine side chain reaching the bonding Asparagine side chain. To remedy this, we introduced a new round of mutants substituting the Leucine residue for an Alanine, whilst preserving the prior mutant motif of Lys/Asn/Asp/Tyr. Similarly, we identified the Valine 541 residue as a steric issue for our Lysine in the IgE CH3 Domain, and thus substituted it for an Alanine. Additionally, our supervisor Dr U Schwarz-Linek suggested that our recombinant Tyrosine residues could sterically interfere with the conserved Tryptophan residue in IgG CH3 Domains (Trp 71 in Mouse, Trp 417 in Human), and potentially that we should alter the site of Tyrosine placement in the IgE CH3 Domain to 511 to give the Tryosine side chain slightly more intramolecular space. Dr Schwarz-Linek also pointed out that the inability of our E. coli BL21(DE3) chassis to reliably form disulfide bonds could have a detrimental effect on protein folding, so we introduced a mutant substituting the relevant Cysteine residues for Valines in our three CH3 Domains. The concluding mutations can be seen below in table 3.
| CH3 Domain | First Mutant | Second Mutant | Third Mutant | Fourth Mutant |
|---|---|---|---|---|
| Human IgG4 (PDB ID: 4B53) | T350K, L365Y, W381D, F423N | T350K, L365Y, W381D, F423N, L441A | T350K, L365Y, W381D, F423N, L441A, W417Y | T350K, L365Y, W381D, F423N, L441A, W417Y, C367V, C425V |
| Mouse IgG (PDB ID: 1CQK) | T11K, L26F, W42D, F84N | T11K, L26F, W42D, F84N, L102A | T11K, L26F, W42D, F84N, L102A, W78Y | T11K, L26F, W42D, F84N, L102A, W78Y, C28V, C86V |
| Human IgE (PDB ID: 5MOL) | A447K, L462Y, W478D, F522N | A447K, L462Y, W478D, F522N, V541A | A447K, L462Y, W478D, F522N, L462V, V541A, V511Y | A447K, L462Y, W478D, F522N, L462V, V541A, V511Y, C464V, C524V |
Table 3: Concluding mutations for antibody CH3 domains following the rational design process of Kwon et al (2017)1. The first mutant is that predicted by the Kwon method, the second, third, and fourth mutants intend to deal with steric issues around the first mutant.
Inducing Isopeptide Bond Formation via Substitution
Our systematic ‘isopeptide hunt’ of the Protein Data Bank (PDB) returned a list of proteins with a double or triple salt bridge motif: Lysines, Arginines, Aspartates, and Glutamates, all held within hydrogen-bonding distance of one another (figure 2). We reasoned that appropriate substitution of one of these residues for an inert amino acid could induce spontaneous formation of an isopeptide bond within the remaining residues. To test this idea, we attempted to edit one of the proteins from this list – the Beta-glucosidase A from the thermophile Thermatoga maritima.
Figure 2: Triple salt bridge motif in the beta-glucosidase from Thermatoga maritima (PDB ID: 2J77). Hydrogen bonds, forming salt bridges between residues, are shown as yellow dashed lines. Figure made using PyMol.
To induce an isopeptide bond, we created the substitutions R371M and R413M, with the intention that Methionine would occupy a similar space in the protein to the Arginine side chain but not form Hydrogen bonds with any residues, preventing salt bridge formation. We also introduced the mutant R371M, R413Y to shield the bond from water with a Tyrosine.
Design of mOrange+SpyTag (BBa_K2929004)
As our part improvement, we proposed codon-optimising the fluorescent protein BBa_E2050 mOrange for Escherichia coli, previously codon-optimised for Yeast. Thematic to our project, we also proposed adding the biobrick BBa_K1159201 ‘SpyTag’ to mOrange to create a part with improved functionality, as the SpyTag would allow ligation to any SpyCatcher-conjugated part via an isopeptide bond6. We conjugated SpyTag to mOrange via a 2xGGSG linker then checked the tertiary structure using our RGN structural predictor. This predicted that SpyTag would be held distal to the mOrange protein (see figure 3), and thus that the SpyTag should be able to ligate with any SpyCatcher presented to the part, and that the SpyTag should not impact the fluorescence of the part and its performance as a reporter.
Figure 3: mOrange+SpyTag tertiary structure as predicted by our RGN tertiary structure predictor. SpyTag is shown in blue, and is held distal to mOrange (green). Image made in PyMol.
We created the part by joining the sequences in silico and ordering the full-length part from IDT, using IDT’s codon-optimisation tool to optimise the part for E. coli.
References
- Kwon, H., Young, P.G., Squire, C.J., and Baker, E.N (2017) – ‘Engineering a Lys-Asn isopeptide bond into an immunoglobulin-like protein domain enhances its stability’ Scientific Reports 7, e42753 (https://doi.org/10.1038/srep42753)
- Holm L. (2019) – ‘Benchmarking fold detection by DaliLite v.5’ Bioinformatics btz536 (https://doi.org/10.1093/bioinformatics/btz536)
- McNicholas, S., Potterton, E., Wilson, K.S., and Noble, M.E.M. (2011) – ‘Presenting your structures: the CCP4mg molecular-graphics software’ Acta Cryst. D67, pp. 386-394 (https://doi.org/10.1107/S0907444911007281)
- Hagan, R.M., Bjornsson, R., MacMahon, S.A., Schomburg, B., Braithwaite, B., Buhl, M., Naismith, J.H., and Schwarz-Linek, U. (2010) – ‘NMR Spectroscopic and Theoretical Analysis of a Spontaneously Formed Lys-Asp Isopeptide Bond’ Angew Chem Int Ed Engl 49(45), pp. 8421–8425 (https://doi.org/10.1002/anie.201004340)
- Hu, X., Hu, H., Melvin, J.A., Clancy, K.W., McCafferty, D.G., and Yang, W. (2010) – ‘Autocatalytic Intramolecular Isopeptide Bond Formation in Gram-Positive Bacterial Pili: A QM/MM Simulation’ J Am Chem Soc 133(3), pp. 478-85 (https://doi.org/10.1021/ja107513t)
- Zakeri, B., Fierer, J.O., Celik, E., Chittock, E.C., Schwarz-Linek, U., Moy, V.T., and Howarth, M. (2012) – ‘Peptide tag forming a rapid covalent bond to a protein, through engineering a bacterial adhesin’ PNAS 109(12), E690-E697 (https://doi.org/10.1073/pnas.1115485109)