VHH CDR-H3 conformation is determined by VH germline usage

Abstract

VHHs or nanobodies are single antigen binding domains originating from camelid heavy-chain antibodies. They are used as diagnostic and research tools and in a variety of therapeutic molecules. Analyzing variable domain structures from llama and alpaca we found that VHHs can be classified into two large structural clusters based on their CDR-H3 conformation. Extended CDR-H3 loops protrude into the solvent, whereas kinked CDR-H3 loops fold back onto framework regions. Both major families have distinct properties in terms of their CDR-H3 secondary structure, how their CDR-H3 interacts with the framework region and how they bind to antigens. We show that the CDR-H3 conformation of VHHs correlates with the germline from which the antibodies are derived: IGHV3-3 derived antibodies almost exclusively adopt a kinked CDR-H3 conformation while the CDR-H3 adopts an extended structure in most IGHV3S53 derived antibodies. We do not observe any bias stemming from V(D)J recombination in llama immune repertoires, suggesting that the correlation is the result of selection processes during B-cell development. Our findings demonstrate a previously undescribed impact of germline usage on antigen interaction and contribute to a better understanding on how properties of the antibody framework shape the immune repertoire.

Fig. 1. Clustering of the CDR-H3 loop of VHH antibodies based on the conformation of the C-terminal stem of the loop.

a Representative conformation of the C-terminal stem of CDR-H3 for extended (top, teal) and kinked (bottom, red) structures are shown. The angles α101 and τ101 which are used for CDR-H3 loop classification are highlighted. b α101 and τ101 angle distributions in a unique set of VHH structures from llama and alpaca (n = 385 VHH structures). Structures highlighted in blue were classified as having an extended CDR-H3 conformation and structures highlighted in red a kinked CDR-H3 conformation. c Aligned CDR-H3 regions which were clustered as extended (top, teal) and as kinked (bottom, red). Structures that were removed from the subsequent analysis are colored in gray (see main text for details).

Fig. 2. Sequence and structural differences in the CDR-H3 and framework of the two structural CDR-H3 clusters.

a Contact map between CDR-H3 and FWR2 in extended CDR-H3 structures (n = 100). The blue heatmap intensity represents the fraction of structures where a particular pair of residues interact with each other (see main text for details). FWR2 and CDR-H3 positions were numbered using the Chothia numbering scheme, except for CDR positions 95 to 100x which were aligned from the C-terminus to reveal conserved interactions at the C-terminus of CDR-H3. The logo plots at the X and Y-axis of the heatmap were generated from the sequences used in the contact analysis. b Contact map as in a for structures carrying kinked CDR-H3 conformation (n = 215). c Cartoon representation of a typical extended CDR-H3 VHH structure (PDB code 4LDE.A) with key residues in CDR-H3 and FWR shown in stick representation. d Cartoon representation of a typical kinked CDR-H3 VHH structure (PDB code 3K7U.A) with key residues in CDR-H3 and FWR shown as in stick representation. e Cartoon representation of a typical kinked CDR-H3 VHH structure (PDB code 1U0Q.A) carrying a short 3–10 helix in CDR-H3; key residues in CDR-H3 and FWR are shown in stick representation.

Fig. 3. The solvent-exposed framework 2 region of antibodies with an extended CDR-H3 loop is frequently involved in antigen binding.

a Boxplot of the relative solvent accessible area of FWR2 position 37, 45, and 47 in VHH structures with extended (teal, n = 100) or kinked CDR-H3 (red, n = 215) (Wilcoxon rank test adjusted p-values, pos 37: 7.7e-32, pos 45: 1.7e-5, pos47: 1.7e-21). b Percent of VHH structures with a kinked (red, n = 162) and extended (teal, n = 86) CDR-H3 which use the different regions (FWR1-4, CDR-H1-3) of the VHH domain for interaction with their respective antigen. Regions that show a significant difference based on a Chi-squared test are labeled (p-values CDR1: 1.7e-3, FWR2: 5.7e-5, FWR4: 4.2e-7).

Fig. 4. The two CDR-H3-based structural clusters differ in their HV germline usage.

a Top 7 most frequent VH germline usage in the dataset of unique VHH structures (n = 385) used in this study. b α101 and τ101 angle distributions in a unique set of VHH structures (n = 106) derived from the IGHV3S53 germline. c α101 and τ101 angle distributions in a unique set of VHH structures (n = 175) derived from the IGHV3-3 germline. d Top 7 most frequent VH germlines used in the combined repertoire dataset of the two sequenced llama samples. e CDR-H3 amino acid length distribution in the combined repertoire of llama 1 and 2 for antibodies either derived from IGHV3-3 (n = 13462 sequences, blue) or IGHV3S53 (n = 18769 sequences, orange) germlines (Wilcoxon rank test adjusted p-value 2e-16). f α101 and τ101 angles predicted by Alphafold for 23 VHH sequences derived from the IGHV3-3 germline (blue) and 19 VHH sequences derived from the IGHV3S53 germline (orange).

Fig. 5. VDJ junction properties of IGHV3-3 and IGHV3S53-derived VHH antibodies.

a IGHD segment usage in the combined repertoire of llama 1 and 2 for antibodies either derived from IGHV3-3 (blue) or IGHV3S53 (orange) germlines. b IGHJ segment usage in the combined repertoire of llama 1 and 2 for antibodies either derived from IGHV3-3 (blue) or IGHV3S53 (orange) germlines. c Llama IgG2 distribution in the combined repertoire of llama 1 and 2 for antibodies either derived from IGHV3-3 (blue) or IGHV3S53 (orange) germlines. d The histogram depicts the distribution of the most 3’ prime nucleotide position in the VH segment of antibodies derived from an IGHV3-3 (upper panel) and IGHV3S53 germline (lower panel). e Nucleotide length distribution of NP nucleotides and D-segment of antibodies either derived from IGHV3-3 (n = 10304 sequences, blue) or IGHV3S53 (n = 14451 sequences, orange) VH germlines (Wilcoxon rank test adjusted p-value 2e-16). f The histogram depicts the distribution of the most 5’ prime nucleotide position in the IGHJ4 segment of antibodies derived from IGHV3-3 (upper panel) and IGHV3S53 germlines (lower panel).

Fig. 6. Framework 2 residue composition at positions 37 and 47 impact stability and CDR-H3 conformation.

a Changes in melting temperature (ΔT_M) between the wildtype version of selected VHHs and their respective F37Y/F47Y mutant version. PDB code and respective chains are listed on the Y-axis. The standard deviation of three technical replicates is represented using an error bar. b Scatter plot between change in melting temperature (ΔT_M) and change in CDR-H3 conformation (root-mean-square deviation; RMSD) as predicted by Omegafold for the VHH molecules shown in a.