Repertoire Analysis of Antibody CDR-H3 Loops Suggests Affinity Maturation Does Not Typically Result in Rigidification

Antibodies can rapidly evolve in specific response to antigens. Affinity maturation drives this evolution through cycles of mutation and selection leading to enhanced antibody specificity and affinity. Elucidating the biophysical mechanisms that underlie affinity maturation is fundamental to understanding B-cell immunity. An emergent hypothesis is that affinity maturation reduces the conformational flexibility of the antibody’s antigen-binding paratope to minimize entropic losses incurred upon binding. In recent years, computational and experimental approaches have tested this hypothesis on a small number of antibodies, often observing a decrease in the flexibility of the complementarity determining region (CDR) loops that typically comprise the paratope and in particular the CDR-H3 loop, which contributes a plurality of antigen contacts. However, there were a few exceptions and previous studies were limited to a small handful of cases. Here, we determined the structural flexibility of the CDR-H3 loop for thousands of recent homology models of the human peripheral blood cell antibody repertoire using rigidity theory. We found no clear delineation in the flexibility of naïve and antigen-experienced antibodies. To account for possible sources of error, we additionally analyzed hundreds of human and mouse antibodies in the Protein Data Bank through both rigidity theory and B-factor analysis. By both metrics, we observed only a slight decrease in the CDR-H3 loop flexibility when comparing affinity matured antibodies to naïve antibodies, and the decrease was not as drastic as previously reported. Further analysis, incorporating molecular dynamics simulations, revealed a spectrum of changes in flexibility. Our results suggest that rigidification may be just one of many biophysical mechanisms for increasing affinity.

Supplementary Data

Supplementary Command Lines
Rosetta version 2017.26-dev59567 was used for all simulations. [1][2][3] Antibody Fv regions were relaxed with the following command and options:

Sequences Used to Model Naïve-Reverted Antibodies
Mature sequences were aligned to germline V-genes as described in the methods. Additionally, sequences were aligned to germline J-genes using IMGT/DomainGapAlign, which yields germline alignments for both V-and J-genes. For example, the alignment of the variable region of 1T2Q can be extracted from: http://www.imgt.org/3Dstructure-DB/cgi/details.cgi?pdbcode=1t2q. The germline sequence was used when possible. The alignments are shown below with the mature sequence above and naive below.

Comparison of Flexibility Calculations Across Ensemble Generation Methods
In this work, we have considered multiple ensemble generation methods in conjunction with FIRST-PG analysis to determine the flexibility of CDR-H3 loops. Of all methods used in this paper to generate structural ensembles, only MD simulations have previously been coupled with flexibility analysis [8][9][10][11] . Rosetta FastRelax, KIC, and RosettaAntibody have not been used previously for flexibility analysis. MD simulations permit for fluctuations between low-energy states and variations in hydrogen bonding networks, effectively capturing the "flickering" nature of hydrogen bonds. The Rosetta-based methods consider hydrogen-bonding energy, but do not involve dynamic motion in the same way as MD simulations. The Rosetta FastRelax protocol generates ensembles representative of the local energy minimum through side-chain repacking and gradient-based energy minimization, so flexibility analysis of these ensembles should be comparable to flexibility analysis of crystal structures. Rosetta KIC on the other hand generates ensembles of low-energy CDR-H3 loop conformations by de novo modeling of the CDR-H3 loop. RosettaAntibody generates ensembles of low-energy antibody conformations, building on KIC motions through additional VH-VL docking. We compared FIRST-PG calculations on Rosetta FastRelax, RosettaAntibody, and MD ensembles, for three well-studied antibodies, excluding KIC ensembles from analysis because they are effectively superseded by RosettaAntibody ensembles. Qualitatively, the FIRST-PG results agree for all methods for 48G7 and the anti-fluorescein antibody. The Rosetta FastRelax results differ from the other methods for the anti-influenza, with the naïve showing significantly more rigidity. This is most likely due to the difference in quality between the crystal structures. Quantitatively, the DAUC values for RosettaAntibody and MD ensembles for all three antibody pairs compare well (Supplemental Table 2). Additionally, we compared only RosettaAntibody and MD ensembles for three naïvereverted/mature antibody pairs, where we found that the DAUC values roughly agree for two out of three antibody pairs. Taken together, these results indicate that flexibility analyses on RosettaAntibody homology model ensembles are similar to analyses on MD ensembles.   Figure 15. CDR-H3 loop B-factor z-scores for antigen-bound and free crystal structures of the catalytic antibody AZ-28 reveal no significant difference between the naïve and mature antibodies.

Supplementary Tables
Supplementary Table 1. List of antibodies analyzed in this study. The following 922 antibodies were studied (attached separately).

Supplementary Table 2. Rigidity changes according to several methods. Changes in the rigidity of the 48G7 antibody CDR-H3 loop according to several methods. Unbound is denoted by (U) and bound is denoted by (B). A positive number indicates an increase in rigidity upon affinity maturation.
Changes for B-factors are calculated as the difference in the average CDR-H3 loop B-factor between the naïve and mature crystal structure: = naive ))))))) − mature ))))))))) ± 0 naive 2 + mature 2 . Changes in FIRST-PG are calculated as the percent change between the AUC of the CDR-H3 melting curve for naïve and mature antibodies: . Finally, changes in MD RMSD or RMSF are calculated as the difference in average CDR-H3 loop RMSF or RMSD between the MD simulations of the naïve and mature antibodies: = naive ))))))) − mature ))))))))) ± 0 naive  Stopped-flow fluorescence measurements on three antibodies reveal binding kinetics with multiple phases indicating that "ligand binding involved isomerization, as well as associative steps." The experimentally characterized antibodies were mature, but the authors speculate that "antibodies in the primary repertoire may be more prone to isomerism" and "affinity maturation in such cases may include mutaitons leading to a more favorable isomeric equilibrium." 20 1 Patten et al. 1996 Science To our knowledge, the first published suggestion of rigidification of the CDR H3 upon maturation, based on studies of the esterolytic antibody 48G7, "affinity maturation appears to play a conformational role, either in reorganizing the active site geometry or limiting sidechain and backbone flexibility of the germline antibody." But no direct evidence is presented. 18 2  Science This paper reports crystal structures for hapten bound/unbound and naïve/mature 48G7 antibodies. Comparison of naïve/mature structural rearrangements upon binding reveals reduced CDR H3 motion of the mature antibody. The authors conclude, "The end result of these somatic mutations is a combining site with improved complementarity to hapten … which, in contrast to the germline antibody, binds hapten in a pre-organized fashion." 37 Chong et al. 1999 PNAS 500 ps MD simulation on 48G7 antibody with hapten present found higher RMSFs in the "belly" atoms from the naïve than the mature antibody. 16  All experiments show that the naïve antibody exhibits more conformational heterogeneity than the mature antibody. MD shows that the naïve motion is primarily in the L1, L2, and H3 CDRs. Therefore, the conclusion is that evolution optimizes a binding site for a specific molecule by preconfiguration. 43 11 Thorpe et al. 2007 PNAS A follow-up paper on #10, with MD simulations of several (varying in maturation) bound/unbound 4-4-20 antibody-antigen complexes. Analysis shows larger fluctuations in the germline unbound simulation than the bound. Calculations of the enthalpies and entropies show a larger change entropy (of binding) upon maturation than in enthalpy (of binding). 44 12 Thielges et al. 2008 Biochemistry This paper analyzes the thermodynamics and dynamics of six antifluorescien antibodies (including 4-4-20). ITC is used to measure enthalpies and entropies of binding. 3PEPS and transient grading (TG) experiments are used to measure motions on the femto-, pico-, and nano-second timescales. Experiments show antibodies net favorable dG binding is due in some cases primarily to a favorable enthalpy change, but one case due to a favorable entropy change. The authors conclude that the immune system can generate antibodies with a variety of dynamics, from rigid antibodies capable of lock-and-key binding to flexible ones capable of induce-fit or conformationalselection binding. 45 13 Babor et al. 2008 Proteins In this paper, the authors use Rosetta Design on antibody crystal structures to recover CDR H3 sequences. When the design is constrained to multiple structures, then naïve sequence is more likely to be recovered than the mature sequence. This is taken to indicate that the naïve sequence is optimal for conformational flexibility. 46 14 Wong et al. 2010 Proteins In this paper, four catalytic antibodies and their naïve equivalents are modeled using MD simulations. Flexibility was assessed by comparing alpha-carbon B-factors for the six CDR loops. In three of the four studied antibodies, a loss of flexibility in residues contacting antigen was observed following maturation. 47 15 Adhikary et al.

JBC
This paper is similar to #12, in that a panel of antibodies was evolved against a chromophore (MPTS) and they were studied by ITC, 3PEPS, and TG. The results are similar to #12, in that the antibodies had varying dynamics and recognized MPTS through multiple mechanisms. The authors conclude that antibodies are initial dynamic with the potential to recognize multiple targets, but eventually are tailored to a single target by evolution. 28 16 Schmidt et al. 2013 PNAS This paper involves crystallization, MD simulation, and SPR studies of broadly neutralizing influenza virus antibodies. The authors show with ~30 microsecond MD simulations that the naïve antibody CDR H3 loop is rarely in the bound conformation, whereas the mature antibody CDR H3 loops occupy the bound conformation between 20-70% of the time. SPR shows that ~66% of the improvement in KD can be attributed to a 10-fold decrease in dissociation rate upon affinity maturation. The authors collectively interpret the results to indicate that "increased conformational restriction of the CDR H3 has been the principle consequence of affinity maturation." 48 17 Willis et al. 2013 PLoS Comp. Bio.
Similar to #13, this paper uses multi/single-state design to assess the structural preference (sequence recovery) of naïve/mature sequences. The authors design on multiple structures originating from the same VH germline, diverging from the previous study. Their results indicate that multi-state design is more likely to recapitulate the naïve sequence whereas single-state design is more likely to recapitulate the mature sequence. The conclusion is that germline sequences possess high conformational flexibility. 49 18 Adhikary et al. 2015 Biochemistry Similar to #12 & #15, this paper uses ITC, 3PEPS, crystallography, and ELISA to analyze the thermodynamics and dynamics of anti-MPTS antibodies. The authors selectively isolate three antibodies with varying specificity for MPTS and other proteins. The authors find that the antibody with the few somatic mutations is the most polyspecific, but also the least dynamic. On the other hand, the antibody with the highest affinity for MPTS is the most dynamic. The authors conclude that affinity maturation can have divergent effects on dynamics towards antigen recognition. 8 19 Li et al. 2015 PLoS Comp. Bio.
The authors assess the flexibility of three antibodies (anti-fluorescein, anti-CD3, 48G7) using MD to generate ensembles and a distance constraint model to evaluate flexibility. The authors note a significant amount of rigidity increases in the CDR H3 loop and flexibility increases in the CDR L2 loop. They believe these effects are compensatory. 50 20 Davenport et al. 2016 Structure This paper studies the effects of SHM on the dynamics of three anti-HIV antibodies using HXMS. The authors find that most stabilization occurred in the CDR L2, H2, and FW3. This contradicts previous studies. The authors rationalize the contradiction as arising due to the relative complexity of HIV antigen versus the previous studied antigen. 51 Supplementary