Team:Queens Canada/Model
Overview
Our project focused heavily on the ability to link recombinant antibody fragments to fluorescent protein; however, not all antibody fragments used had predicted structures. Moreover, fusing proteins to antibodies may affect the ability of the antibody to bind to its antigen – the fusion should occur far from the complimentary determining regions (CDR). For the proteins with known structures, we modelled our antibody with GFP to determine optimal linkage sites, based on distance from the CDRs. The antibody with an unknown structure was submitted for homology structure determination and the resulting protein model was used to determine optimal linkage sites that will not interfere with the CDRs of the antibody.
ScFv-CDR Determination and Modeling
The anti-THC ScFv had the complimentary regions previously determined; however, there is currently no available PDB file for this protein (1). In order to attach a fluorescent protein to this antibody fragment, we had to be sure that it is unlikely to interfere with the complimentary determining regions (CDRs) of the antibody; therefore, we modelled the predicted structure.
ABodyBuilder was used for structure determination, which is a structure determination software designed specifically for nanobodies (2). Structures are based on orientation prediction, CDR modelling, and side chain prediction, and results are given a confidence score based on the root-mean square deviation threshold.
The Following ScFv protein sequences were submitted for the ScFv.
HC: QVQLQESGPGLVKPSETLSLTCTVSGGSISSGYYWGWIRQPPGKGLEWIGSIYHSGSTYYNPSLKSRVTISVDTSKNQFSLKLSSVTAA
DTAVYYCARGSAKRAVKWGQGTLVTVSSG
LC: ALQTVVTQEPSFSVSPGGTVTLTCGLSSGSVSTSYYPSWYQQTPGQAPRTLIYSTNTRSSGVPDRFSGSILGNKAALTITGAQADDESDY
YCVLYMGSGVVFGGGTKLTVLG
From this model, the confidence score for the VH and VL frameworks were 1.00, and 0.88, indicating high confidence in the model (Fig. 1). Although it was noted that the CDR H3 and CRD L3 regions had the least similarity to known structures.
Although the confidence in the CDR H3 and L3 regions was not as high, all predicted CDRs were identical to previously predicted CDRs (1) (we could not use their predicted model, as they did not release their PDB file). The only difference between ABodyBuilder and Dr. Brennan’s model was in CDR H2 – the ABodyBuilder CDR H2 was longer on the N-term by two amino acids (Fig. 2). However, due to the near identical CDR regions determined independently from Dr. Brenna, we believe this to be an accurate model of the anti-THC antibody fragment.
From the ABodyBuilder model we used the generated PDB file to determine optimal linkage to a fluorescent protein (Fig. 3A). Hence, the optimal binding sites to connect the fluorescent protein to are the N-termini of the LC and HC. However, the N-terminus of the HC is occupied, as it links to the C-terminal of the LC, via a GGGGSGGGGSGGS linker. Therefore, the only optimal linkage point for the fluorescent protein is the N-terminal of the LC (Fig. 3B).
The model of the ScFv was later verified, as it was shown that the fluorescent antibody generated from this model was able to bind to THC (Fig. 4). For more details about the THC binding assay please visit the Results page.
Fab-Model
Unlike the ScFv, the Fab used in this experiment did have a known crystal structure (PDB: 3ls4) (3). Based on the structure of the anti-THC Fab we designed a linker large enough to bind the fluorescent protein. We chose to attach the fluorescent protein to the N-terminal of the heavy chain, rather than the light chain, as the heavy chain is usually attached to another domain (Fig. 5A). Additionally, we aimed to develop a truncated version of the Fab, by using the fluorescent protein as a linker between the N-terminal of the light chain and the C-terminal of the heavy chain (Fig. 5B). It has been recorded that using a fluorescent protein as a linker between the heavy and light chains can increase the solubility of the recombinantly expressed antibody (4).
Summary
The structure of the anti-THC ScFv has not been determined; hence, we modelled the ScFv to determine optimal linkage to a fluorescent protein. ABodyBuilder predicted the structure of the anti-THC ScFv, based on template selection, orientation prediction, complementary-determining region (CDR) loop modeling, and side chain prediction (2). The root-mean-square deviation (RMSD) for the predicted heavy and light chain model are 1.00, and 0.88, respectively, indicating high model confidence (Fig. 1). Moreover, the predicted CDR regions determined by ABodyBuilder agree with the previously predicted CDRs by the researchers who characterized the antibody (1). Additionally, the model indicated that the N and C-terminal of the light chain were too close to the binding site; however, the C-terminal of the heavy chain was suitable for linkage to a fluorescent protein (Fig. 2). Therefore, the fluorescent protein was linked to the ScFv on the C-terminal of the heavy chain.
References
1. Brennan, J. (2005) The production of recombinant single chain antibody fragments for the detection of illicit drug residues. doctoral thesis, Dublin City University, [online] http://doras.dcu.ie/17319/ (Accessed March 12, 2019)
2. ABodyBuilder: Automated antibody structure prediction with data–driven accuracy estimation: mAbs: Vol 8, No 7 [online] https://www.tandfonline.com/doi/full/10.1080/19420862.2016.1205773?scroll=top&needAccess=true (Accessed October 9, 2019)
3. Niemi, M. H., Turunen, L., Pulli, T., Nevanen, T. K., Höyhtyä, M., Söderlund, H., Rouvinen, J., and Takkinen, K. (2010) A Structural Insight into the Molecular Recognition of a (−)-Δ9-Tetrahydrocannabinol and the Development of a Sensitive, One-Step, Homogeneous Immunocomplex-Based Assay for Its Detection. Journal of Molecular Biology. 400, 803–814
4. Markiv, A., Beatson, R., Burchell, J., Durvasula, R. V., and Kang, A. S. (2011) Expression of recombinant multi-coloured fluorescent antibodies in gor -/trxB- E. coli cytoplasm. BMC Biotechnol. 11, 117
