Allow me to present our recently accepted paper: Comparative analysis of the CDR loops of antigen receptors, to appear in Frontiers in Immunology [1]. In the blog post I will give a quick five-minute summary of the key messages in this work.
Our immune system recognises foreign molecules, known as antigens, and removes them. To trigger the immune system, immune cells need to recognise self from non-self antigens. This recognition is mainly carried out by B-cells and T-cells.
B-cell and T-cell receptors, hereafter generalised as antibodies and TCRs, are encoded by similar genetic mechanisms. Both receptors undergo V(D)J recombinations to achieve high variability in their antigen-binding sites. These genes encode for heavy and light chains in antibodies, and correspondingly β and α chains in αβ TCRs. Structurally, both receptors adopt a β-sandwich configuration. Given how similar they are, however, antibodies can bind to an almost infinite range of antigens while TCRs can only target peptide antigens presented by the major histocompatibility complex. We then wonder, if we can analyse the binding sites and find out why they have different targets.
In this work, we only focus on their complementarity determining regions (CDRs) that form the majority of both of their binding sites. Five of the six CDRs adopt a limited number of backbone conformations, known as the ′canonical classes′. The remaining CDR (β3 in TCRs and H3 in antibodies) is more structurally diverse.
We began with an update to the current definition of canonical classes. We then compared the canonical CDRs of the two receptors. For CDRs of the corresponding types (i.e. CDRβ1/H1 and CDRα1/L1), they tend to have different lengths. This makes it a bit more difficult for them to be structurally similar even if we use a length-independent structural clustering algorithm [2]. We proved this by clustering both antibody and TCR CDRs. They separate into different structural clusters, except in two of them where we find both types of CDRs. When we looked at their sequence patterns (see an example of one of the two clusters in Figure a), TCR clearly use a distinguished pattern to form these structures. If we mixed both types of sequences together, the strong sequence pattern becomes less prominent. To quantify this difference, we one-hot-encoded the sequences, and observed that they mostly separate into distinct regions by Principal component analysis (Figure B). These results point towards:
- CDR from TCR and antibodies don’t share similar sequence lengths.
- When they do, they don’t resemble each other structurally.
- And when they do, they don’t share similar sequence patterns.
The second set of analysis we did was on the the anchor regions of the β3/H3 loops. Previous analysis showed that in H3, although the loop conformations were highly variable, their base regions were mainly kinked rather than extended. We ran the same analysis on the base region of β3, and found the intriguing figure of 97.5% of them adopting an extended base region.
Finally, and possibly the key observation we made in this paper, we analysed all the CDR sequences that were solved multiple times and have multiple example structures. There has always been an interesting discussion on whether loops are flexible. What we found is that, in antibodies, 8% of the loops that could have adopted multiple conformations, actually did so. On the contrary, nearly 20% of their TCR counterparts adopt multiple conformations. But given the lack of structural data for TCRs, it might not be sufficient to challenge the application of canonical form modelling in TCRs. We need more data!
The differences between the main binding site components (CDRs) in these antigen receptors may point towards their different roles in the immune system. We see an increasing interest in leveraging TCR components to design antibodies that can target intracellular antigens, and grafting antibody features to enhance specificity in TCRs for therapeutic treatment. Our study highlighted important structural characteristics that should be considered whilst making these design decisions. A fuller understanding of the structures can improve design of the binding.
Reference:
- Wong, W.K., Leem, J. and Deane, C.M., 2019. Comparative analysis of the CDR loops of antigen receptors. Frontiers in Immunology, doi: 10.3389/fimmu.2019.02454.
- Nowak, J., Baker, T., Georges, G., Kelm, S., Klostermann, S., Shi, J., Sridharan, S. and Deane, C.M., 2016. Length-independent structural similarities enrich the antibody CDR canonical class model. MAbs, 8(4), pp. 751-760.
