Leveraging NGS sequence data: we highlight insights from Dr Isidro Hotzel at Department of Antibody Engineering, Genentech
Olivia Siviter highlights research presented by Dr Isidro Hotzel from the Department of Antibody Engineering at Genentech at Antibody Engineering & Therapeutics Europe 2021.
In his presentation, Dr Isidro Hotzel from the Department of Antibody Engineering at Genentech discusses how to leverage NGS sequence information for antibody optimization, antibody discovery and functional understanding of antibody repertoires.
Variable regions of antibodies are generated by the recombination of germline segments in VDJ recombination. As there is no variability in the H1 and H2 regions of the V segment, most diversity is between antibodies is in the CDR H3 region.
The light chains of antibodies only consist of a V and J segment, so there is less variability between light chains when compared to the heavy chains. Once B cells encounter an antigen, they undergo clonal expansion and somatic mutation.
Before leveraging NGS sequence information, hybridoma technology was used to mine clones for discovery. However, this process only mines a limited number of clones making the process of antibody discovery much slower.
Dr Hotzel’s lab set out to use deep sequencing sampling to enhance the sampling for the identification of clones with better properties, specifically affinity.
In this combinational method, variant antibody sequence identification is performed in parallel with hybridoma or B cell cloning, meaning that there is no added time for affinity optimization during the process of discovery and no need for library generation or selection.
As clones of the same lineage will have the same V and J segments, they can select for clones with the same V and J segments and then at least 60% amino acid identity in the CDR H3 segment.
This is a relatively low threshold to ensure that all possible clones are included. This produces NGS read of the whole repertoire in a sample which allows for the identification of unique CDR and vernier residues in a clonotype.
Variants in the clonotype can then be selected and affinities/binding kinetics can be measured. This enables the identification of clones with the highest affinity. Light chains and heavy chains can be combined to produce clones with the highest affinities.
Dr Hotzel’s lab found that by using this method, they could find clones with up to 60x improved affinity demonstrating that this data mining technique is a robust discovery technique.
"The light chains of antibodies only consist of a V and J segment, so there is less variability between light chains when compared to the heavy chains."
In traditional NGS, a sequence yield of millions is obtained, however the chain pairing information is lost, meaning that it cannot enable antibody discovery and functional repertoire characterization.
When you combine B cell lineage mining and NGS, the sequence yield is much less (thousands compared to millions). However, the chain pairing information in retained resulting in larger date sets that can lead to antibody discovery.
This method can also be extended for functional repertoire characterization using a barcoded gel-beads in emulsion technology.
This technology is high throughput with ~30 000 cells sequenced per run and ~93-98% pairing accuracy making large-scale functional repertoire characterization easier.
Off the back of the development of this system, Dr Hotzel’s lab set out to see if it was possible to expand repertoire mining beyond the confines of B cell lineages using sequence information, as B-cell lineage mining only allows for mining within the chosen lineage.
The disadvantage of this is that antibodies will have the same binding geometry and bind to the same epitope.
The occurrence of antibody convergences was investigated to discover the existence of parallel lineages. Parallel lineages are antibodies that bind to an antigen with the same Vh/VL germline segment pair but a different CDR H3 region.
To understand whether parallel lineages are a rare event or a predictable repertoire property, antibody competition assays were performed. 18 overlapping epitope groups in 41 mAbs in rat anti OVA sub-panel were discovered.
Of these 18 overlapping epitope groups, 16 shared the same Vh/Vl germline segments. It was found that epitope restriction by Vh/Vl germline segment pairing is highly predictable; 71/74 anti-OVA antibodies had predictable epitopes based off their Vh/Vl segments. Similar results were also found with anti-lysosome antibodies.
This led to the question of whether CDR H3 played a major role in binding of parallel lineages.
To test this, Dr Hotzel’s lab randomized CDR H3 and Her2 specific variants were selected by phage display. They found the epitope specificity was retained in all the hits. This shows that convergence was not dependent on any special binding property of these clones and that parallel lineages are a higher level of antibody repertoire organization.
The parallel lineage framework can be used to recognise functional redundancies from sequence information alone. NGS can be performed NGS across lineages to understand CDR3 H3 regions with the best fit for affinity optimization.
