To study structural changes that occur in Abs upon Ag binding, we systematically compared free and bound structures of all 141 crystal structures of the 49 Abs that were solved in these two forms. We found that many structural changes occur far from the Ag binding site. Some of them may constitute a mechanism for the recently suggested allosteric effects in Abs. Within the binding site itself, CDR-H3 is the only element that shows significant binding-related conformational changes; however, this occurs in only one third of the Abs. Beyond the binding site, Ag binding is associated with changes in the relative orientation of the H and L chains in both the variable and constant domains. An even larger change occurs in the elbow angle between the variable and the constant domains, and it is significantly larger for binding of big Ags than for binding of small ones. The most consistent and substantial conformational changes occur in a loop in the H chain constant domain. This loop is implicated in the interaction between the H and L chains, is often intrinsically disordered, and is involved in complement binding. Hence, we suggest that it may have a role in Ab function. These findings provide structural insight into the recently proposed allosteric effects in Abs.

The inherent flexibility of proteins is essential for their function, allowing them to adopt new conformations and, in turn, bind to distinct ligands (13). Experimental and structural studies suggest that Abs are no different, showing some flexibility upon Ag binding (4). This flexibility was suggested to be essential to their ability to bind multiple Ags (3). However, although the existence of conformational changes in the Ag binding site has been widely recognized, their role in the adaptive immune system has not been structurally elucidated. Furthermore, understanding these conformational changes should help to improve Ab modeling, docking, and engineering (59).

Until recently, the variable and constant domains were considered functionally independent, where Ab affinity and specificity are determined by the variable domains, and Ab isotype and effector function are mediated by the constant domains (10, 11). Thus, most of the Ag binding-related structural changes were expected to occur at the variable domains. However, recent reports provide some evidence for binding-related allosteric effects in Abs. It was demonstrated that Ag binding, as well as the structure of the variable domains, may be influenced by changes in the constant domains (1124). For example, Pritsch et al. (13) demonstrated that two human mAbs sharing identical variable domains, but expressing different isotypes, bind tubulin with significantly different affinities. Because the differences observed were found at the Fab level, the investigators suggested a role for the constant heavy-1 (CH1) domain in shaping the Ag binding site. Janda and Casadevall (11) used circular dichroism spectra to analyze four mAb isotypes of the 3E5 family that share identical variable domains. They found that the different isotypes undergo different structural changes upon binding to a common Ag, providing evidence for structural cross-talk between the constant and the variable regions. Tudor et al. (24) even suggested that the epitopes recognized by two anti–HIV-1 IgG1 and IgA2 Abs with identical variable domains are only partially overlapping. Furthermore, several studies suggested that long-distance structural changes may be transferred in the other direction as well, from the variable to the constant domain, potentially influencing effector activation (10, 2527). For example, Oda et al. (25) showed that the binding of staphylococcal protein A (SPA) or streptococcal protein G to the constant domain was inhibited by hapten binding in several mAbs. Results of isothermal titration calorimetry also revealed that the Ka for the interaction of SPA with IgG2b was decreased by the addition of hapten. Because SPA and streptococcal protein G are known to bind to IgG CH domains, such as CH1, CH2, and CH3, the investigators concluded that signals resulting from the hapten binding induced a conformational change at these constant domains. A different example was provided by Horgan et al. (27), who observed differences in complement activation of two Abs that differ only in their variable-heavy (VH) domain. A possible mechanism for such observations was suggested by Torres and Casadevall (10), in which electrostatic and hydrophobic interactions resulting from differences in the microenvironment of CH domains (e.g., pH, ionic strength) may affect the Ag binding site. Additionally, the arrangement of the Fab constant domains relative to the variable domains and to each other may increase the probability of an appropriate VH–variable light (VL) relative orientation (28), which, in turn, can shape the Ag binding site (4, 29, 30).

Early analyses during the 1990s and the beginning of the millennium attempted to characterize structural changes that occur in Abs upon Ag binding (8, 2835). The results of these studies, which typically relied on only a few Abs, revealed a large variation in the binding-related conformational changes in individual Abs. This included reports of very small overall changes, side-chain movements, large rearrangement of one or more of the CDRs, change in the VH–VL relative orientation, and, in some cases, combinations of some or all of the above. Recent studies used newer structure-comparison methods (4, 36) and used the growing number of available structures (3740) to perform comparative research to identify new and more general characteristics of Ab structures. However, many of these studies did not compare free and bound structures of the same Ab (4, 37, 40), focused only on the VH–VL relative orientation (4, 37, 38, 40), or included only a limited number of structures (4, 36, 39). None of these studies elaborated on the Fab constant domains.

In this study, we describe a comprehensive systematic structural comparison of all Abs whose crystal structure has been experimentally determined both in free and in Ag-bound forms. We characterize changes at the residue level in both the variable and the constant domains, as well as changes in the heavy–light and variable–constant relative orientations. We also compare different bound structures of the same Ab to determine whether the observed changes could be conclusively ascribed to Ag binding rather than to intrinsic flexibility.

The Protein Data Bank (PDB) ID of all crystal structures containing Abs, as well as the Ab chain IDs in these PDB files, were identified using International Immunogenetics Information System/3Dstructure-DB, version 4.5.0 (41, 42). For each PDB file, only one biological unit (the first one that appears in “REMARK 350”) was considered. Heavy and light labels were assigned to the Ab chains by running BLAST (43) against an example sequence of one H chain and one L chain and selecting the one with the higher E value. The Ab in each structure was labeled either “free” (if no other chain in the same biological unit exists within 6 Å of the Ab variable domain) or “bound” (otherwise). Structures containing an Fc region, structures of germline Abs, and scFv structures were ignored. We also discarded structures of Abs bound to a non-amino acid Ag, to a peptide Ag with less than five amino acids, or to an Ag that is an Ab itself. To identify different crystal structures of the same Ab, the sequences of all Ab chains were clustered using BLASTCLUST (44) (performed separately for H and L chains), requiring 100% sequence identity and 95% coverage. Coverage was required in only one sequence, so that two structures of the same Ab, one containing both variable and constant domains and the other containing only the variable domain, would be grouped together. Two structures were considered as being of the same Ab if both of their chains (heavy and light) shared 100% sequence identity. The final dataset contained only Abs that were found to have at least one free and one bound structure. The PDB IDs of the Abs in this dataset are shown in Table I. For example, cluster number 7 contains one free structure and six bound ones. Thus, for this Ab, we measured free–bound changes over the 6 free–bound pairs and bound–bound changes over the 15 bound–bound pairs.

Table I.
The dataset
Cluster NumberPDB CodeH ChainL ChainFree/BoundAg ChainsaConstant Domainb
1vfa Free – − 
1vfb Bound − 
1fgn Free – 
1ahw Bound 
1mqk Free – − 
1ar1 Bound AB − 
1qle Bound ABCD − 
3ehb Bound AB − 
3hb3 Bound AB − 
1ay1 Free – 
1bgx Bound 
1bey Free – 
1ce1 Bound 
1bvl Free – − 
1bvk Bound − 
1cfq Free – 
1cfn Bound 
1cfs Bound 
1cft Bound 
1hh6 Bound 
1hh9 Bound 
1hi6 Bound 
1ck0 Free – 
2ck0 Bound 
3ck0 Bound 
1cr9 Free – 
1cu4 Bound 
10 1dqm Free – 
10 1dqq Free – 
10 1dqj Bound 
10 1nby Bound 
10 1nbz Bound 
11 1e6o Free – 
11 1e6j Bound 
12 1f8t Free – 
12 1f90 Bound 
13 1fvc Free – − 
13 1n8z Bound 
14 1gig Free – 
14 2vir Bound 
14 2vis Bound 
14 2vit Bound 
15 1hil Free – 
15 1him Bound 
15 1hin Bound 
15 1ifh Bound 
16 1kcu Free – 
16 1kc5 Bound 
17 1kcv Free – 
17 1kcs Bound 
18 1mf2 Free – 
18 2hrp Bound 
19 1nlb Free – 
19 1n64 Bound 
20 1mim Free – 
20 3iu3 Bound 
21 1mlb Free – 
21 1mlc Bound 
22 3d69 Free – 
22 1nl0 Bound 
23 1oaq Free – − 
23 1ocw Free – − 
23 1oaz Bound − 
24 1om3 Free – 
24 2oqj Bound 
25 1qbl Free – 
25 1wej Bound 
26 1rhh Free – 
26 2b4c Bound CG 
27 1rz8 Free – 
27 1rzj Bound CG 
27 1rzk Bound CG 
27 1yyl Bound MG 
27 1yym Bound MG 
27 2i5y Bound MG 
27 2i60 Bound MG 
27 2nxy Bound AB 
27 2nxz Bound AB 
27 2ny0 Bound AB 
27 2ny1 Bound AB 
27 2ny2 Bound AB 
27 2ny3 Bound AB 
27 2ny4 Bound AB 
27 2ny5 Bound CG 
27 2ny6 Bound AB 
28 1u6a Free – 
28 3hi1 Bound 
29 2zkh Free – 
29 1v7m Bound 
29 1v7n Bound 
30 1yy8 Free – 
30 1yy9 Bound 
31 2g75 Free – 
31 2dd8 Bound 
32 2eh7 Free – 
32 2eh8 Bound 
33 2fat Free – 
33 2fd6 Bound AU 
33 3bt2 Bound ABU 
34 2fjf Free – 
34 2fjg Bound WV 
35 2pr4 Free – 
35 3d0l Bound 
35 3d0v Bound 
35 3idg Bound 
35 3idi Bound 
35 3idj Bound 
35 3idm Bound 
35 3idn Bound 
36 2vxu Free – 
36 2vxt Bound 
37 3cvi Free – 
37 3cvh Bound ABC 
38 3eo9 Free – 
38 3eoa Bound 
38 3eob Bound 
39 3eyo Free – 
39 3eyf Bound 
40 3gje Free – 
40 3gjf Bound ABC 
41 3hi5 Free – 
41 3hi6 Bound 
42 3pp3 Free – 
42 3pp4 Bound 
43 1bbd Free – 
43 1a3r Bound 
44 1igf Free – 
44 2igf Bound 
45 1l7i Free – 
45 1s78 Bound 
46 1mnu Free – 
46 1mpa Bound 
46 2mpa Bound 
47 2gsg Free – − 
47 2otu Bound − 
47 2otw Bound − 
48 2hkh Free – 
48 2hkf Bound 
49 3bkc Free – 
49 3bkm Free – 
49 3bkj Bound 
Cluster NumberPDB CodeH ChainL ChainFree/BoundAg ChainsaConstant Domainb
1vfa Free – − 
1vfb Bound − 
1fgn Free – 
1ahw Bound 
1mqk Free – − 
1ar1 Bound AB − 
1qle Bound ABCD − 
3ehb Bound AB − 
3hb3 Bound AB − 
1ay1 Free – 
1bgx Bound 
1bey Free – 
1ce1 Bound 
1bvl Free – − 
1bvk Bound − 
1cfq Free – 
1cfn Bound 
1cfs Bound 
1cft Bound 
1hh6 Bound 
1hh9 Bound 
1hi6 Bound 
1ck0 Free – 
2ck0 Bound 
3ck0 Bound 
1cr9 Free – 
1cu4 Bound 
10 1dqm Free – 
10 1dqq Free – 
10 1dqj Bound 
10 1nby Bound 
10 1nbz Bound 
11 1e6o Free – 
11 1e6j Bound 
12 1f8t Free – 
12 1f90 Bound 
13 1fvc Free – − 
13 1n8z Bound 
14 1gig Free – 
14 2vir Bound 
14 2vis Bound 
14 2vit Bound 
15 1hil Free – 
15 1him Bound 
15 1hin Bound 
15 1ifh Bound 
16 1kcu Free – 
16 1kc5 Bound 
17 1kcv Free – 
17 1kcs Bound 
18 1mf2 Free – 
18 2hrp Bound 
19 1nlb Free – 
19 1n64 Bound 
20 1mim Free – 
20 3iu3 Bound 
21 1mlb Free – 
21 1mlc Bound 
22 3d69 Free – 
22 1nl0 Bound 
23 1oaq Free – − 
23 1ocw Free – − 
23 1oaz Bound − 
24 1om3 Free – 
24 2oqj Bound 
25 1qbl Free – 
25 1wej Bound 
26 1rhh Free – 
26 2b4c Bound CG 
27 1rz8 Free – 
27 1rzj Bound CG 
27 1rzk Bound CG 
27 1yyl Bound MG 
27 1yym Bound MG 
27 2i5y Bound MG 
27 2i60 Bound MG 
27 2nxy Bound AB 
27 2nxz Bound AB 
27 2ny0 Bound AB 
27 2ny1 Bound AB 
27 2ny2 Bound AB 
27 2ny3 Bound AB 
27 2ny4 Bound AB 
27 2ny5 Bound CG 
27 2ny6 Bound AB 
28 1u6a Free – 
28 3hi1 Bound 
29 2zkh Free – 
29 1v7m Bound 
29 1v7n Bound 
30 1yy8 Free – 
30 1yy9 Bound 
31 2g75 Free – 
31 2dd8 Bound 
32 2eh7 Free – 
32 2eh8 Bound 
33 2fat Free – 
33 2fd6 Bound AU 
33 3bt2 Bound ABU 
34 2fjf Free – 
34 2fjg Bound WV 
35 2pr4 Free – 
35 3d0l Bound 
35 3d0v Bound 
35 3idg Bound 
35 3idi Bound 
35 3idj Bound 
35 3idm Bound 
35 3idn Bound 
36 2vxu Free – 
36 2vxt Bound 
37 3cvi Free – 
37 3cvh Bound ABC 
38 3eo9 Free – 
38 3eoa Bound 
38 3eob Bound 
39 3eyo Free – 
39 3eyf Bound 
40 3gje Free – 
40 3gjf Bound ABC 
41 3hi5 Free – 
41 3hi6 Bound 
42 3pp3 Free – 
42 3pp4 Bound 
43 1bbd Free – 
43 1a3r Bound 
44 1igf Free – 
44 2igf Bound 
45 1l7i Free – 
45 1s78 Bound 
46 1mnu Free – 
46 1mpa Bound 
46 2mpa Bound 
47 2gsg Free – − 
47 2otu Bound − 
47 2otw Bound − 
48 2hkh Free – 
48 2hkf Bound 
49 3bkc Free – 
49 3bkm Free – 
49 3bkj Bound 
a

Structures that do not contain Ag chains are labeled “−”.

b

Structures that contain or do not contain the constant domain are labeled “+” or “−,” respectively.

The analysis described below was performed separately for the L and H chains. Structural features for each residue of each protein in the dataset were compared between each pair of structures (i.e., between one free and one bound structure or between two bound structures) of the same Ab. The structural features were calculated as follows:

  1. Solvent accessibility was calculated with STRIDE (45) and DSSP (46). For bound structures, the Ag was not included in the calculation so as to avoid changes that are a direct result of Ag binding rather than a conformational change. The change in solvent accessibility was then calculated as the absolute value of the difference between the solvent accessibility of the two structures. Positions for which this change was not consistent between DSSP and STRIDE were ignored.

  2. The number of interchain contacts was calculated for each residue by counting the number of amino acids that it contacts in the opposite chain. Two amino acids are considered to be in contact if any of their respective atoms are within 6 Å of each other. The change in interchain contacts was calculated as the absolute value of the difference between the number of interchain contacts of the two structures.

  3. Cα and heavy atoms root-mean-square deviation (RMSD) of two compared structures were calculated for each position, following their structure alignment using combinatorial extension (CE) (47). The structure alignment was performed separately for the constant and variable domains.

Next, corresponding positions in different Abs were identified using the nomenclature of the Multiple Structure Alignment (MSTA) of 96 H chains and 98 L chains described previously (48). For structures that were not included in the MSTA, a pair-wise sequence alignment was performed against all Abs in the MSTA using BLAST, and the most similar sequence (highest E value) was identified. The Ab of interest was then structurally aligned to this most closely sequence-related structure using CE, allowing the mapping of its residues to the MSTA.

Finally, the RMSD and the changes in solvent accessibility and interchain contacts were averaged, first over all pairs of either free–bound or bound–bound structures of the same Ab (to avoid bias toward Abs that have more structures than others) and then over all Abs.

The space group in which each structure was solved was extracted from “REMARK 290” in the PDB file. The position-based comparisons described above were then repeated, including only pairs of structures that were solved with the same space group.

The CDRs were defined as described by Ofran et al. (48). The first loop of the constant heavy-1 (CH1-1) domain was defined between positions 159 and 171 in the MSTA of Ofran et al. (48). The RMSD was averaged first over all pairs of either free–bound or bound–bound structures of the same cluster and then over all Abs. For each two segments, the statistical significance of their average RMSD difference was computed using the following reshuffling procedure: all RMSD values (one from each Ab) of the two segments were randomly divided into two new lists (of the same size as the original ones), and the random average difference was calculated by subtracting the average of one list from that of the other. This process was repeated 10,000 times, and the p value was calculated. The RMSD averages of two segments for which p ≤ 0.05 were considered to be significantly different.

To compare the relative orientation of the H chain versus the L chain, we used a strategy similar to that of other investigators (4, 33, 38); we superimposed two equivalent chains of the structures being compared and calculated the Cα RMSD between the two other equivalent chains. Intradomain changes are ignored by reducing the intradomain RMSD. Specifically, for each pair of free–bound or bound–bound structures of the same Ab, the RMSDvariable and RMSDconstant, which estimate the change in the heavy–light relative orientation of the variable or of the constant domain, respectively, were calculated as follows (steps 1 and 2 are also illustrated in Fig. 4):

FIGURE 4.

Schematic illustration of the first two steps in calculating the change in VH–VL relative orientation (see 2Materials and Methods). (A) The two structures are superimposed according to the variable domain of the H chains, and RMSDall is calculated based on the L chain variable domains. (B) The two structures are superimposed according to the variable domain of the L chains, and RMSDintra is calculated based on the L chain variable domains. The PDB IDs of the two structures are 1e6o (green–orange) and 1e6j (blue–red). The H chains are colored green and red, and the L chains are colored orange and blue.

FIGURE 4.

Schematic illustration of the first two steps in calculating the change in VH–VL relative orientation (see 2Materials and Methods). (A) The two structures are superimposed according to the variable domain of the H chains, and RMSDall is calculated based on the L chain variable domains. (B) The two structures are superimposed according to the variable domain of the L chains, and RMSDintra is calculated based on the L chain variable domains. The PDB IDs of the two structures are 1e6o (green–orange) and 1e6j (blue–red). The H chains are colored green and red, and the L chains are colored orange and blue.

Close modal
  1. The heavy variable domains of the two structures were structurally aligned using CE, and the Cα RMSD between the two variable light domains (RMSDall) was calculated.

  2. The light variable domains of the two structures were structurally aligned using CE, and the Cα RMSD between these two variable light domains (RMSDintra) was calculated.

  3. RMSDlightvariable was calculated by subtracting RMSDintra from RMSDall.

  4. Similarly, RMSDheavyvariable was calculated by applying steps 1–3 to the other chain.

  5. The RMSDvariable is the average of RMSDlightvariable and RMSDheavyvariable.

RMSDconstant was calculated by applying steps 1–5 to the constant domains instead of the variable domains. The RMSDvariable and the RMSDconstant were averaged first over all pairs of either free–bound or bound–bound structures of the same Ab, and then over all Abs. The p values were calculated as described above.

For each pair of free–bound or bound–bound structures of the same Ab, RMSDheavy and RMSDlight, which estimate the change in the variable–constant relative orientation of the H chain and the L chain, respectively, as well as the corresponding p values, were calculated in a similar way to the heavy–light relative orientation described above.

A residue was considered disordered if its coordinates were missing from the PDB file. The CH1-1 loop of a structure was considered disordered if at least one of its residues is disordered.

The Cα B-factors of all residues were extracted from the PDB files. The averaged B-factor of the CH1-1 loop was calculated by averaging the B-factors over the loop positions (159–171), over all structures of the same Ab, and then over all Abs. For comparison, the B-factor was averaged over all other Fab positions (excluding the CH1-1 loop), as well.

The sequence of all alleles of human IgG genes were downloaded from the International Immunogenetics Information System Web site (49) and aligned using ClustalW (50). The conservation scores for each position were calculated using JALVIEW (51, 52).

We retrieved all Abs with experimentally determined three-dimensional structures in both free and bound forms. We considered only pairs of structures that share 100% sequence identity in both the H and L chains. Table I lists 141 PDB structures representing 49 different Abs that make up our dataset. Thirteen of the 49 Abs had more than one bound structure, allowing the comparison of two bound structures. This bound–bound comparison provides a control set that allows us to determine whether an observed change is Ag-binding related: a conformational change may be considered as binding related only if it is observed between free and bound structures but not between different bound structures of the same Ab. Overall, our dataset consists of 94 free–bound and 162 bound–bound pairs of structures of the same Ab. Twenty-two of the Abs were bound to peptide Ags (≤33 aa), and the other 27 were bound to proteins (the shortest of which was 104 aa). The sequence identity of the Ags bound to the same Ab in different structures varies in peptide Ags from 100% (e.g., 1hin versus 1him) to 0% (e.g., 1cfs versus 1hh9). In protein Ags, the sequence identity varies from 100% (e.g., 1yyl versus 1yym) to 83% in the entire Ag but 95% in the epitope (e.g., 1rzk versus 2ny5). For 6 of the 49 Abs, at least one structure consists of only the variable domain, which precludes their use in calculating conformational changes in the constant domains and variable–constant relative orientation changes.

To identify the structural changes that occur at the residue level between free and bound structures of the same Ab, we calculated the RMSD (once for Cα only and once for all heavy atoms), the change in solvent accessibility, and the change in the number of interchain contacts for each position in each of these pairs of structures. The results from structurally equivalent positions of different free–bound pairs of the same Ab and of different Abs were averaged. When possible, the same calculation was performed for pairs of bound structures of the same Ab to characterize the structural changes as related to Ag binding or not. Fig. 1 shows the observed changes, at the residue level, for free–bound (blue) and bound–bound (red) comparisons for the following features: Cα RMSD (Fig. 1A), heavy-atoms RMSD (Fig. 1B), change in solvent accessibility (Fig. 1C), and change in the number of contacts with the other chain (Fig. 1D). In this representation, two regions clearly stand out: CDR-H3 and the first loop of the CH1 domain (CH1-1 loop) that is located between positions 159 and 171 and shown in Fig. 2. For all features considered, the differences between free and bound in these regions were substantial. The conformational change in CDR-H3 appears only when comparing free and bound structures and not when comparing two bound structures, suggesting that the conformational change is directly related to Ag binding. Changes in the CH1-1 loop are observed in both free–bound and bound–bound comparisons, suggesting, that in this case, at least part of the change may be ascribed to factors other than Ag binding. The five other CDRs (CDR-H1, CDR-H2, and CDR-L1–L3) do not exhibit significant changes in any of the parameters relative to the frameworks (FRs) or the constant domains. The overall similarity in the patterns of the Cα and heavy-atoms RMSD (Fig. 1A versus Fig. 1B) indicates that no specific region (including the CDRs) is characterized by movements of the side chains alone. It also indicates that free–bound changes are not enriched in backbone or side chain rearrangements compared with bound–bound changes, and vice versa.

FIGURE 1.

Position-based comparison of free–bound and bound–bound structures. Cα RMSD (A), heavy-atoms RMSD (B), change in solvent accessibility (C), and change in contacts with the other chain (D). The H chain residues are presented in the left panels, and the L chain residues are presented in the right panels. Free–bound comparisons are shown in blue, and bound–bound comparisons are shown in red. The CDRs are presented as plus signs, and all other residues are depicted as dots. Positions for which the data originated from fewer than three Abs are not shown, as well as the elbow region connecting the variable and constant domains and the structure edges. (E) The structure of BH-151 Ab (PDB ID: 1eo8) is colored according to the average free-bound Cα RMSD values shown in (A), from blue (low RMSD values) to red (high RMSD values). Regions for which the RMSD was not calculated are colored black. β-strands and loops are presented as wide and narrow ribbons, respectively.

FIGURE 1.

Position-based comparison of free–bound and bound–bound structures. Cα RMSD (A), heavy-atoms RMSD (B), change in solvent accessibility (C), and change in contacts with the other chain (D). The H chain residues are presented in the left panels, and the L chain residues are presented in the right panels. Free–bound comparisons are shown in blue, and bound–bound comparisons are shown in red. The CDRs are presented as plus signs, and all other residues are depicted as dots. Positions for which the data originated from fewer than three Abs are not shown, as well as the elbow region connecting the variable and constant domains and the structure edges. (E) The structure of BH-151 Ab (PDB ID: 1eo8) is colored according to the average free-bound Cα RMSD values shown in (A), from blue (low RMSD values) to red (high RMSD values). Regions for which the RMSD was not calculated are colored black. β-strands and loops are presented as wide and narrow ribbons, respectively.

Close modal
FIGURE 2.

The CH1-1 loop in the Ab scaffold. (A) HyHEL-63 Ab CH1 and CL constant domains (PDB ID: 1dqm). The CH1 and CL domains are colored cyan and light green, respectively. The CH1-1 loop residues and the CL residues within 5 Å of the CH1-1 loop are presented as sticks. The Cys of the CH1-1 loop and the Cys of the CL domain connected by an S-S bond are colored by element. (B) The entire Ab scaffold (PDB ID: 1igy). The H and L chains are colored blue and orange, respectively. The CH1-1 loop is presented as a space-filling representation and is colored light green.

FIGURE 2.

The CH1-1 loop in the Ab scaffold. (A) HyHEL-63 Ab CH1 and CL constant domains (PDB ID: 1dqm). The CH1 and CL domains are colored cyan and light green, respectively. The CH1-1 loop residues and the CL residues within 5 Å of the CH1-1 loop are presented as sticks. The Cys of the CH1-1 loop and the Cys of the CL domain connected by an S-S bond are colored by element. (B) The entire Ab scaffold (PDB ID: 1igy). The H and L chains are colored blue and orange, respectively. The CH1-1 loop is presented as a space-filling representation and is colored light green.

Close modal

In the variable domain, six loops point toward the constant domain, opposite the CDRs (three on each chain). Two of them, the second on each chain (located at FR-2), show larger changes compared with the other four (Fig. 1A–D: residues 51–55 and 55–59 in the H and L chains, respectively, and Fig. 1E: VH-2 and VL-2 loops). These two loops interact with each other, and together with CDR-H3 and CDR-L3 they constitute a major part of the heavy–light interface (32); thus, their conformational changes may affect the heavy–light relative orientation and, as a result, the Ag binding site itself.

Overall, some periodicity can be observed in Fig. 1, including some peaks that do not correspond to the CDRs or to the loops discussed above. This periodicity is a result of the secondary structure of the Ab scaffold, with low-flexibility regions corresponding to β-strands and peaks corresponding to loops. For all CDRs other than CDR-H3, the flexibility, as reflected by the peaks, is similar to that seen in nonbinding loops.

In almost all positions, the RMSD between free and bound structures is greater than the RMSD between two bound structures. A possible explanation we explored is that most free–bound pairs (74%) were solved in different space groups, whereas most of the bound–bound pairs (58%) were solved in identical space groups (Supplemental Fig. 1A). Thus, one might expect that free–bound pairs will more often have different crystal contacts, which may be translated into structural diversity. To test this hypothesis, we reanalyzed the Cα RMSD only for pairs of free–bound and bound–bound structures solved in the same space group. However, the results of this analysis indicate that the overall trend remains, with free–bound changes clearly greater than bound–bound ones, in nearly all Fab positions (Supplemental Fig. 1B). Thus, these differences could not be ascribed solely to differences in space groups.

To quantify the conformational changes that occur in specific segments of interest (i.e., the CDRs and the CH1-1 loop) relative to each other and to the rest of the Fab, we recalculated the Cα RMSD of free–bound and bound–bound pairs for each segment (instead of for each position) separately. We also calculated the Cα RMSD of a “baseline” (i.e., all residues of the Fab chains, both variable and constant domains) excluding the CDRs and the CH1-1 loop. As shown in Fig. 3A, all CDRs exhibited greater changes in free–bound comparisons than in bound–bound comparisons (p < 0.001, p < 0.001, p < 0.001, p ≤ 0.003, p ≤ 0.008, and p ≤ 0.002, for CDR-H1, -H2, -H3, -L1, -L2, and -L3, respectively). As suggested in the position-based analysis, the baselines of both H and L chains differ more when comparing free with bound than when comparing two bound structures (p < 0.001 and p ≤ 0.001 for the baseline of the H and L chain, respectively). The free–bound average RMSDs of the CDRs and the baselines are still significantly higher than are the bound–bound ones when including only pairs of structures solved with the same space group (p ≤ 0.005 for the CDRs, p ≤ 0.024 for the baselines).

FIGURE 3.

Intra- and interdomain changes. (A) Segment-based comparisons. Dark and light gray bars represent the Cα RMSD of free–bound and bound–bound pairs, respectively. Standard errors are shown as error bars. (B) Average heavy–light relative orientation change, in the variable and constant domains, of free–bound (dark gray bars) and bound–bound (light gray bars) pairs. (C) Average variable–constant relative orientation change, in the H and L chains, of free–bound (dark gray bars) and bound–bound (light gray bars) pairs. Standard errors are shown as error bars.

FIGURE 3.

Intra- and interdomain changes. (A) Segment-based comparisons. Dark and light gray bars represent the Cα RMSD of free–bound and bound–bound pairs, respectively. Standard errors are shown as error bars. (B) Average heavy–light relative orientation change, in the variable and constant domains, of free–bound (dark gray bars) and bound–bound (light gray bars) pairs. (C) Average variable–constant relative orientation change, in the H and L chains, of free–bound (dark gray bars) and bound–bound (light gray bars) pairs. Standard errors are shown as error bars.

Close modal

The CH1-1 loop is the region in the Fab structure that exhibits the largest conformational change, with an average Cα RMSD of 1.8 Å (Fig. 3A). This value is significantly higher even than that of CDR-H3 (p ≤ 0.038). In addition, a change >1Å in this loop is common for 63% of the Abs (Supplemental Fig. 2). The average RMSD of free–bound is higher than that of bound–bound. As indicated by the error bars in Fig. 3A, the variability of the RMSDs in this loop in different Abs is high. The difference between free–bound and bound–bound has a p value ≤ 0.045. However, when including only pairs of structures solved with the same space group, the p value increases to ≤0.357. Thus, it is impossible to establish that the changes of this loop are related to Ag binding. The flexibility of CH-1 is also reflected by the observation that its bound–bound average RMSD is significantly higher than that of the other CDRs and the baseline (p ≤ 0.002). We also found that 37% of the structures in our dataset have intrinsic disorder in this region (Supplemental Table IA) and that the averaged B-factor of this loop is high compared with the rest of the Fab (Supplemental Table IB).

Of the CDRs, CDR-H3 has the highest free–bound average RMSD (1.30 Å), which is significantly greater than the baseline (p < 0.001). Nonetheless, the free–bound RMSD of CDR-H3 is >1 Å in only 37% of the Abs (Supplemental Fig. 2) (i.e., a large conformational change of this CDR upon Ag binding is not a common feature of most Abs). However, when such changes in the conformation of CDR-H3 were observed, they occurred between the free form and any of the bound structures (as opposed to between the free structure and only some of the bound ones). In contrast to the free–bound change, the bound–bound change in CDR-H3 is not significantly higher than that of the other CDRs and the baseline (p ≤ 0.312), suggesting that when conformational changes occur in CDR-H3 they are related to Ag binding.

The free–bound average RMSDs of CDR-H1, -H2, -L1, and -L3 are ∼0.5 Å and are not significantly different from those of the baselines (p ≤ 0.47, p ≤ 0.41, p ≤ 0.27, and p ≤ 0.24, respectively). Nevertheless, there are some specific examples in which the RMSD for these CDRs is high. Specifically, the RMSD of CDR-H1, CDR-H2, CDR-L1, and CDR-L3 was >1 Å in 12, 14, 4, and 4% of the Abs, respectively (Supplemental Fig. 2). The free–bound average RMSD of CDR-L2, 0.35 Å (Fig. 3A), is significantly lower than that of all other CDRs and even than the free–bound RMSD of the baseline (p < 0.001).

Next, we characterized free–bound and bound–bound changes in the relative orientation of the H and L chains (see Materials and Methods and Fig. 4). Fig. 3B shows that the change in the heavy–light relative orientation of both variable and constant domains is significantly higher in free–bound comparisons than in bound–bound comparisons (p < 0.001 and p ≤ 0.001 for the variable and constant domains, respectively). This observation suggests that Ag binding affects the relative orientation of the H and L chains both in the variable domain and in the constant domain. The change in the VH–VL relative orientation is significantly higher than that of CH1-constant light (CL) when comparing free and bound structures (p < 0.001) but not when comparing two bound structures (p ≤ 0.07). Thus, we conclude that the higher conformational change in VH–VL (relative to that of CH1-CL) is the result of Ag binding. However, this effect is relatively minor, even in the variable domain, with an average RMSD of 0.58 Å and a maximum value of 1.76 Å.

In a similar way, we also characterized the relative orientation of the variable versus the constant domain. Fig. 2C shows that the variable–constant relative orientation change in free–bound comparisons is remarkable (average RMSD of 3.51 and 3.60 Å, and a maximum value of 17.88 and 19.01 Å for the H and L chain, respectively), compared with the change observed in heavy–light relative orientation. It seems that the change in variable–constant relative orientation is a result of the Ag binding, because the average RMSD of the free–bound comparisons is significantly higher than that of the bound–bound comparisons (p < 0.001 for both H and L chains). Variable–constant relative orientation change is similar in the H and L chains.

Because the Ag binding may induce such a significant change in the variable–constant relative orientation, we hypothesized that this effect may depend on the Ag size. Fig. 5 shows that, indeed, the average RMSD of variable–constant relative orientation induced by protein Ags is significantly higher than the change induced by peptide Ags (p ≤ 0.002). Although the type of Ag cannot be trivially used for predicting the amount of change in the variable–constant relative orientation, because some of the protein Ags still show a very small change (e.g., 0.30 Å between 3gje and 3gjf), dramatic changes can almost be excluded for peptide Ags, because 18 of these 21 Ags do not induce a change that is >1.74 Å. Surprisingly, a similar relationship between the Ag size and the heavy–light relative orientation was not found, nor were such relationships found between the Ag size and the CDR-H3 RMSD or between the Ag size and the CH1-1 loop RMSD.

FIGURE 5.

The variable–constant relative orientation change in peptide versus protein Ags. Standard error bars are shown.

FIGURE 5.

The variable–constant relative orientation change in peptide versus protein Ags. Standard error bars are shown.

Close modal

In 1993, when the first comparative studies of Abs appeared (32, 34), the structures of only 33 Abs were available. With the structures of almost 1400 Abs available in the PDB, we attempted to reassess binding-related conformational changes. Of the six CDRs, CDR-H3 shows the largest conformational change upon binding the Ag, as was previously suggested (30, 32, 36). It has the highest number of contacts with the Ag (53) and is believed to play a key role in Ag recognition (54). It is also the only CDR that shows a considerable conformational change (≥0.6Å Cα RMSD) when comparing free and bound germline Abs (39). It was suggested (39) that this flexibility of CDR-H3 in germline Abs enables the binding of different Ags, and that during affinity maturation, it becomes more rigid with a conformation that is optimal for a specific Ag. Our results indicate that in many of the Abs, CDR-H3 shows only minor conformational changes between free and bound forms. However, despite the rigidification of the Ab during affinity maturation, CDR-H3 still shows substantial structural changes in 37% of the Abs and an average Cα RMSD of 1.3 Å between free and bound structures (and sometimes this RMSD is >3 Å). None of the other CDRs displayed a conformational change greater than the changes observed for the rest of the Fab structure (the FRs and the constant domains). CDR-L2 shows the smallest structural change upon Ag binding. Notably, this change is significantly lower than that of the other CDRs and is even smaller than the change in the FR and the constant domain (the baseline). It is not the shortest CDR, but it has the lowest number of contacts with the Ag (53), a low number of mutations during affinity maturation (55), and low structural diversity in different Abs (56). These observations may suggest that the conformational change that a CDR undergoes is related to its role in Ag binding.

Structural changes upon Ag binding also occur in the relative orientation of the Fab domains. Such interdomain movements were suggested to affect protein functions in other systems as well (38). More than 20 y ago, Colman et al. (57, 58) suggested that the Ag binding site may be shaped by repositioning of the VH domain with respect to the VL. This suggestion, further supported by more recent studies (4, 29, 30, 38), highlights the significance of understanding the relative orientation of these two chains. We found that the relative orientation of the H–L chains changes more when comparing free–bound structures than when comparing bound–bound structures, in both the variable and the constant domains. These findings are in agreement with the idea that the VH–VL relative orientation contributes to Ag binding, but they also suggest that the CH1-CL relative orientation shows a similar (yet less prominent) effect. The observation that the loops in the heavy and light FR-2s, which play an important role in the heavy–light interaction (32), show some conformational change, further demonstrates the relationship between the VH–VL interface and Ag binding. With this in mind, the actual change in free–bound relative orientation of VH versus VL is still limited, in agreement with a small rotation <3° observed previously (33). The Ag-binding–related changes observed in VH–VL relative orientation are smaller than the difference in the related orientation observed between different Abs (38).

The relative orientation of the variable–constant domains in both the H and the L chains undergoes a more prominent change than does the VH–VL domains. As in the case of heavy–light, the change observed in free–bound is greater than the change observed in bound–bound, indicating that these changes are related to Ag binding. Changes in the elbow angle as a result of Ag binding were also suggested by a molecular dynamics simulation (59). The large change in the variable–constant relative orientation does not necessarily support the already disproved theory of an “open” conformation for a free Ab and a “closed” one for the bound form (30); rather, it demonstrates the high flexibility of the Ab, introduced by the elbow region between the variable and the constant domains. This flexibility, which results in a more significant conformational change in the elbow angle than in the VH–VL relative orientation upon Ag binding, was also suggested to occur when comparing different Abs (as opposed to different structures of the same Ab) (32). What could be the role of this flexibility? A change in the variable–constant relative orientation may change the contacts between residues in these two domains (32), and this may change the VH–VL relative orientation, shaping the binding site for Ag binding. Alternatively, Ag binding may cause a change in the VH–VL relative orientation, which, in turn, may change the contacts between residues in the variable–constant interface, resulting in a different elbow angle. This different elbow angle could be translated into a different CH1-CL relative orientation and possibly enable or disable the binding of an effector to the constant domains. These mechanisms suggest that a relationship between the change in elbow angle and Ag binding is plausible, as observed in our results. Additional support for this relationship is the dependence of the elbow angle change on the Ag size: when the Ag is a protein, the average change in the elbow angle between free and bound structures is more than three times the change observed for peptide Ags.

Surprisingly, the region showing the greatest and most consistent conformational change between free and bound structures is the CH1-1 loop, which is located in the constant domain, far from the Ag binding site. Although we were not able to prove that the conformational changes in the CH1-1 loop are related to Ag binding, because substantial conformational changes were observed in this loop in bound–bound comparisons as well, some of its characteristics, as well as several previous results, suggest that it may have a role in Ab function. As shown in Fig. 2A, this loop is part of the interface with the CL domain; in some cases, it is even connected to the CL domain through a disulfide (S-S) bond, thereby having a potential effect on the CH1-CL relative orientation. Furthermore, this loop is close in space to the hinge region connecting the CH1 domain to the Fc (Fig. 2B). Thus, a conformational change in this loop may affect the relative orientation of the CH1 domain versus the Fc, thereby influencing effector binding to the Fc. There are now many examples of Abs with identical variable domains but different isotypes that bind the same Ag with a different affinity or specificity (13, 1524). Some of these studies suggested that the CH1 domain accounts for the observed changes in binding, because it was identified as the only region with sequence diversity between the tested Abs and because a similar effect was observed using only the Fab instead of the entire Ab scaffold (13, 16, 19). Supplemental Fig. 3 shows a multiple sequence alignment of all human IgG CH1 alleles. Sequence diversity within this domain is observed in only two regions, one of which is the CH1-1 loop, suggesting that it may play a role in the functional diversity of Abs. In addition to its flexibility, the CH1-1 loop is also intrinsically disordered, at least in part, in more than one third of the structures in our dataset, which may suggest the existence of a functional binding site (60). Although the majority of Ab effector functions (e.g., complement activation and interaction with FcRs) are mediated by binding of the Fc, it was suggested that complement binding is also mediated by the CH1-1 loop. In particular, C3b was shown to covalently bind Ser132 [Eu numbering (61)], which belongs to the CH1-1 loop, during complement binding to the Ab–Ag complex (62). Another example of the potential role of the CH1 domain in Ab activity is the lack of ability to activate the alternative pathway of complement by IgG molecules with the inter–CH1-CL S-S bond reduced (63). A possible mechanism for the effect of a conformational change in the CH1-1 loop on Ag binding could be a change in the CH1-1 loop microenvironment, as a result of effector binding or a change in the hinge region, which may result in a change to the CH1-CL relative orientation. As discussed above and suggested previously (26, 28, 62), this may lead to changes in the VH–VL relative orientation through the elbow angle and, thus, to changes in the Ag binding site. Such a mechanism is also consistent with a recent analysis that found a large difference in affinity between a Fab and its corresponding Fv fragment (12). Similarly, a signal may be transferred following Ag binding, from the variable, through the variable–constant interface and the elbow angle, to the CH1 domain, and possibly through the hinge region to the Fc domains. Additional studies may shed more light on the nature of this loop and its role, or lack thereof, in the immune response.

It is generally accepted that Abs do not bind Ags in a rigid lock-and-key manner but rather exist in multiple conformational states (8). However, it is unclear whether the Ag binding induces the change or binds to pre-existing conformation [“pre-existing equilibrium” (3, 35)]. Although structural changes between free and bound Abs were commonly associated with an induced-fit mechanism, structural evidence for the pre-existing equilibrium was found for SPE7 Ab (35); two pre-existing conformations of the Ag-free form were crystallized, each conferring a different Ag-binding function. A normal-mode analysis study of two Abs (3), showing that the intrinsic fluctuations of the free conformation correlate with the structural changes observed upon Ag binding, is also in agreement with this hypothesis. Our finding, that some Abs show a significant conformational change upon Ag binding whereas others do not, can be explained by the existence of multiple free conformations in equilibrium, some of which are capable of binding specific Ags. If the free conformation that was crystallized is the same one bound to the complexed Ag, only minor structural changes between the free and bound structures will occur. However, if a different pre-existing free conformation was crystallized, significant changes are expected. Additional analyses of Abs with more than one free structure, as well as bound ones, will further reveal the extent to which this hypothesis is supported.

It is believed that a single Ab may bind multiple Ags, because Abs have a limited repertoire of structures yet can theoretically bind an infinite number of Ags (3). It was suggested that this ability is facilitated by the Ab conformational flexibility (3). Thus, one may expect the similarity of two Ags, and the structural changes that occur in the Ab that binds them both, to be correlated to some extent. Multiple bound structures of the same Ab, with some variability in the Ag epitopes, are required to search for such a correlation. Our dataset includes only two such clusters. Indeed, in one of these clusters, a correlation was observed between epitope sequence similarity and the resulting changes in the Ab structure (data not shown). New structures solved in the coming years are likely to provide additional examples of multiple structures of different Ags bound to the same Ab and will enable a more comprehensive analysis in this regard.

A potential concern when comparing free and bound crystal structures is the existence of possible experimental factors that affect free or bound structures in different ways, which may lead to some artifacts. For example, Fabs tend to crystallize in a similar lattice. Crystallization of Fab–Ag complexes (especially when the Ag is large) is likely to occur in a different lattice. We attempted to address this concern by analyzing only pairs of structures that were solved with the same space group. Notably, comparing only structures that have similar unit cell sizes may have been more accurate; however, there are no complexes of Abs bound to a protein Ag that have a similar unit cell size as the free Ab. An additional potential artifact may result from the fact that, in general, Ab–Ag complexes tend to have more biological impact than do their free counterparts. This may cause crystallographers to invest more effort into trying to solve bound structures, which may result in the use of unusual crystallization conditions. A possible way to avoid such an artifact is the comparison of only pairs of structures that were both solved in common space groups, as suggested previously (64). However, the current dataset does not allow for such analyses because there are hardly any pairs of this sort. Thus, we presented our analysis, keeping in mind these potential biases.

mAbs represent a growing segment of biological drugs for treating a variety of diseases (5). Ab structures, and specifically Ab–Ag complexes, can help to overcome challenges in Ab stabilization, affinity maturation, and humanization, all of which are required for the successful development of therapeutic Abs (7). When a structure of the Ab–Ag complex does not exist, it can be modeled with computational docking tools, relying on the structures of the monomers. However, for an accurate Ab–Ag docking, the structural changes occurring in the Ab and Ag upon binding should be recognized. Indeed, the inherent flexibility of proteins is considered one of the most challenging topics in molecular modeling (9). The comprehensive characterization of free–bound changes in Abs provided in this article may to help improve Ab modeling and Ab–Ag docking algorithms by incorporating the conclusions presented in this study into the modeling techniques and docking algorithms.

We thank Turkan Haliloglu, Sharon Fishman, Anat Burkovitz, Guy Nimrod, Vered Kunik, and Ariel Feiglin for valuable comments and suggestions.

This work was supported by Grant 511/10 from the Israeli Science Foundation.

The online version of this article contains supplemental material.

Abbreviations used in this article:

CE

combinatorial extension

CH1

constant heavy-1

CH1-1

first loop of the constant heavy-1 domain

CL

constant light

FR

framework

MSTA

multiple structure alignment

PDB

Protein Data Bank

RMSD

root-mean-square deviation

SPA

staphylococcal protein A

S-S

disulfide

VH

variable heavy

VL

variable light.

1
Bhalla
J.
,
Storchan
G. B.
,
MacCarthy
C. M.
,
Uversky
V. N.
,
Tcherkasskaya
O.
.
2006
.
Local flexibility in molecular function paradigm.
Mol. Cell. Proteomics
5
:
1212
1223
.
2
Gunasekaran
K.
,
Nussinov
R.
.
2007
.
How different are structurally flexible and rigid binding sites? Sequence and structural features discriminating proteins that do and do not undergo conformational change upon ligand binding.
J. Mol. Biol.
365
:
257
273
.
3
Keskin
O.
2007
.
Binding induced conformational changes of proteins correlate with their intrinsic fluctuations: a case study of antibodies.
BMC Struct. Biol.
7: 31.
4
Pellequer
J. L.
,
Chen
S.
,
Roberts
V. A.
,
Tainer
J. A.
,
Getzoff
E. D.
.
1999
.
Unraveling the effect of changes in conformation and compactness at the antibody V(L)-V(H) interface upon antigen binding.
J. Mol. Recognit.
12
:
267
275
.
5
Caravella
J. A.
,
Wang
D.
,
Glaser
S. M.
,
Lugovskoy
A.
.
2010
.
Structure-guided gesign of antibodies. Curr. Comput. Aided Drug Des.
6
:
128
138
.
6
Sircar
A.
,
Gray
J. J.
.
2010
.
SnugDock: paratope structural optimization during antibody-antigen docking compensates for errors in antibody homology models.
PLOS Comput. Biol.
6
:
e1000644
.
7
Sircar
A.
,
Kim
E. T.
,
Gray
J. J.
.
2009
.
RosettaAntibody: antibody variable region homology modeling server.
Nucleic Acids Res.
37
(
Web Server issue
):
W474–W479
.
8
Stanfield
R. L.
,
Wilson
I. A.
.
1994
.
Antigen-induced conformational changes in antibodies: a problem for structural prediction and design.
Trends Biotechnol.
12
:
275
279
.
9
Spyrakis
F.
,
BidonChanal
A.
,
Barril
X.
,
Luque
F. J.
.
2011
.
Protein flexibility and ligand recognition: challenges for molecular modeling.
Curr. Top. Med. Chem.
11
:
192
210
.
10
Torres
M.
,
Casadevall
A.
.
2008
.
The immunoglobulin constant region contributes to affinity and specificity.
Trends Immunol.
29
:
91
97
.
11
Janda
A.
,
Casadevall
A.
.
2010
.
Circular Dichroism reveals evidence of coupling between immunoglobulin constant and variable region secondary structure.
Mol. Immunol.
47
:
1421
1425
.
12
Adachi
M.
,
Kurihara
Y.
,
Nojima
H.
,
Takeda-Shitaka
M.
,
Kamiya
K.
,
Umeyama
H.
.
2003
.
Interaction between the antigen and antibody is controlled by the constant domains: normal mode dynamics of the HEL-HyHEL-10 complex.
Protein Sci.
12
:
2125
2131
.
13
Pritsch
O.
,
Hudry-Clergeon
G.
,
Buckle
M.
,
Petillot
Y.
,
Bouvet
J. P.
,
Gagnon
J.
,
Dighiero
G.
.
1996
.
Can immunoglobulin C(H)1 constant region domain modulate antigen binding affinity of antibodies?
J. Clin. Invest.
98
:
2235
2243
.
14
Rodrigo
W. W.
,
Block
O. K.
,
Lane
C.
,
Sukupolvi-Petty
S.
,
Goncalvez
A. P.
,
Johnson
S.
,
Diamond
M. S.
,
Lai
C.-J.
,
Rose
R. C.
,
Jin
X.
,
Schlesinger
J. J.
.
2009
.
Dengue virus neutralization is modulated by IgG antibody subclass and Fcgamma receptor subtype.
Virology
394
:
175
182
.
15
Dam
T. K.
,
Torres
M.
,
Brewer
C. F.
,
Casadevall
A.
.
2008
.
Isothermal titration calorimetry reveals differential binding thermodynamics of variable region-identical antibodies differing in constant region for a univalent ligand.
J. Biol. Chem.
283
:
31366
31370
.
16
Torres
M.
,
Fernández-Fuentes
N.
,
Fiser
A.
,
Casadevall
A.
.
2007
.
The immunoglobulin heavy chain constant region affects kinetic and thermodynamic parameters of antibody variable region interactions with antigen.
J. Biol. Chem.
282
:
13917
13927
.
17
McLean
G. R.
,
Torres
M.
,
Elguezabal
N.
,
Nakouzi
A.
,
Casadevall
A.
.
2002
.
Isotype can affect the fine specificity of an antibody for a polysaccharide antigen.
J. Immunol.
169
:
1379
1386
.
18
Torres
M.
,
May
R.
,
Scharff
M. D.
,
Casadevall
A.
.
2005
.
Variable-region-identical antibodies differing in isotype demonstrate differences in fine specificity and idiotype.
J. Immunol.
174
:
2132
2142
.
19
Pritsch
O.
,
Magnac
C.
,
Dumas
G.
,
Bouvet
J. P.
,
Alzari
P.
,
Dighiero
G.
.
2000
.
Can isotype switch modulate antigen-binding affinity and influence clonal selection?
Eur. J. Immunol
.
30
:
3387
3395
.
20
Cooper
L. J.
,
Shikhman
A. R.
,
Glass
D. D.
,
Kangisser
D.
,
Cunningham
M. W.
,
Greenspan
N. S.
.
1993
.
Role of heavy chain constant domains in antibody-antigen interaction. Apparent specificity differences among streptococcal IgG antibodies expressing identical variable domains.
J. Immunol.
150
:
2231
2242
.
21
McCloskey
N.
,
Turner
M. W.
,
Steffner
P.
,
Owens
R.
,
Goldblatt
D.
.
1996
.
Human constant regions influence the antibody binding characteristics of mouse-human chimeric IgG subclasses.
Immunology
88
:
169
173
.
22
Michaelsen
T. E.
,
Ihle
O.
,
Beckstrøm
K. J.
,
Herstad
T. K.
,
Sandin
R. H.
,
Kolberg
J.
,
Aase
A.
.
2003
.
Binding properties and anti-bacterial activities of V-region identical, human IgG and IgM antibodies, against group B Neisseria meningitidis.
Biochem. Soc. Trans.
31
:
1032
1035
.
23
Liu
F.
,
Bergami
P. L.
,
Duval
M.
,
Kuhrt
D.
,
Posner
M.
,
Cavacini
L.
.
2003
.
Expression and functional activity of isotype and subclass switched human monoclonal antibody reactive with the base of the V3 loop of HIV-1 gp120.
AIDS Res. Hum. Retroviruses
19
:
597
607
.
24
Tudor
D.
,
Yu
H.
,
Maupetit
J.
,
Drillet
A. S.
,
Bouceba
T.
,
Schwartz-Cornil
I.
,
Lopalco
L.
,
Tuffery
P.
,
Bomsel
M.
.
2012
.
Isotype modulates epitope specificity, affinity, and antiviral activities of anti-HIV-1 human broadly neutralizing 2F5 antibody.
Proc. Natl. Acad. Sci. USA
109
:
12680
12685
.
25
Oda
M.
,
Kozono
H.
,
Morii
H.
,
Azuma
T.
.
2003
.
Evidence of allosteric conformational changes in the antibody constant region upon antigen binding.
Int. Immunol.
15
:
417
426
.
26
Piekarska
B.
,
Drozd
A.
,
Konieczny
L.
,
Król
M.
,
Jurkowski
W.
,
Roterman
I.
,
Spólnik
P.
,
Stopa
B.
,
Rybarska
J.
.
2006
.
The indirect generation of long-distance structural changes in antibodies upon their binding to antigen.
Chem. Biol. Drug Des.
68
:
276
283
.
27
Horgan
C.
,
Brown
K.
,
Pincus
S. H.
.
1992
.
Effect of H chain V region on complement activation by immobilized immune complexes.
J. Immunol.
149
:
127
135
.
28
Padlan
E. A.
1994
.
Anatomy of the antibody molecule.
Mol. Immunol.
31
:
169
217
.
29
Braden
B. C.
,
Poljak
R. J.
.
1995
.
Structural features of the reactions between antibodies and protein antigens.
FASEB J.
9
:
9
16
.
30
Wilson
I. A.
,
Stanfield
R. L.
.
1994
.
Antibody-antigen interactions: new structures and new conformational changes.
Curr. Opin. Struct. Biol.
4
:
857
867
.
31
Arevalo
J. H.
,
Hassig
C. A.
,
Stura
E. A.
,
Sims
M. J.
,
Taussig
M. J.
,
Wilson
I. A.
.
1994
.
Structural analysis of antibody specificity. Detailed comparison of five Fab′-steroid complexes.
J. Mol. Biol.
241
:
663
690
.
32
Davies
D. R.
,
Chacko
S.
.
1993
.
Antibody Structure.
Acc. Chem. Res.
26
:
421
427
.
33
Li
Y. L.
,
Li
H. M.
,
Smith-Gill
S. J.
,
Mariuzza
R. A.
.
2000
.
Three-dimensional structures of the free and antigen-bound Fab from monoclonal antilysozyme antibody HyHEL-63(,).
Biochemistry
39
:
6296
6309
.
34
Schulze-Gahmen
U.
,
Rini
J. M.
,
Wilson
I. A.
.
1993
.
Detailed analysis of the free and bound conformations of an antibody. X-ray structures of Fab 17/9 and three different Fab-peptide complexes.
J. Mol. Biol.
234
:
1098
1118
.
35
James
L. C.
,
Roversi
P.
,
Tawfik
D. S.
.
2003
.
Antibody multispecificity mediated by conformational diversity.
Science
299
:
1362
1367
.
36
Silverman
B. D.
2007
.
Using molecular principal axes for structural comparison: determining the tertiary changes of a FAB antibody domain induced by antigenic binding.
BMC Struct. Biol.
7:
77
.
37
Abhinandan
K. R.
,
Martin
A. C.
.
2010
.
Analysis and prediction of VH/VL packing in antibodies.
Protein Eng. Des. Sel.
23
:
689
697
.
38
Narayanan
A.
,
Sellers
B. D.
,
Jacobson
M. P.
.
2009
.
Energy-based analysis and prediction of the orientation between light- and heavy-chain antibody variable domains.
J. Mol. Biol.
388
:
941
953
.
39
Babor
M.
,
Kortemme
T.
.
2009
.
Multi-constraint computational design suggests that native sequences of germline antibody H3 loops are nearly optimal for conformational flexibility.
Proteins
75
:
846
858
.
40
Chailyan
A.
,
Marcatili
P.
,
Tramontano
A.
.
2011
.
The association of heavy and light chain variable domains in antibodies: implications for antigen specificity.
FEBS J.
278
:
2858
2866
.
41
Ehrenmann
F.
,
Kaas
Q.
,
Lefranc
M. P.
.
2010
.
IMGT/3Dstructure-DB and IMGT/DomainGapAlign: a database and a tool for immunoglobulins or antibodies, T cell receptors, MHC, IgSF and MhcSF.
Nucleic Acids Res.
38
(
Database issue
):
D301
D307
.
42
Kaas
Q.
,
Ruiz
M.
,
Lefranc
M. P.
.
2004
.
IMGT/3Dstructure-DB and IMGT/StructuralQuery, a database and a tool for immunoglobulin, T cell receptor and MHC structural data.
Nucleic Acids Res.
32
(
Database issue
):
D208
D210
.
43
Altschul
S. F.
,
Madden
T. L.
,
Schäffer
A. A.
,
Zhang
J. H.
,
Zhang
Z.
,
Miller
W.
,
Lipman
D. J.
.
1997
.
Gapped BLAST and PSI-BLAST: a new generation of protein database search programs.
Nucleic Acids Res.
25
:
3389
3402
.
44
Dondoshansky
I.
,
Wolf
Y.
.
2002
.
Blastclust (NCBI Software Development Toolkit).
National Center for Biotechnology Information
,
Bethesda, MD.
45
Frishman
D.
,
Argos
P.
.
1995
.
Knowledge-based protein secondary structure assignment.
Proteins
23
:
566
579
.
46
Kabsch
W.
,
Sander
C.
.
1983
.
Dictionary of protein secondary structure: pattern recognition of hydrogen-bonded and geometrical features.
Biopolymers
22
:
2577
2637
.
47
Shindyalov
I. N.
,
Bourne
P. E.
.
1998
.
Protein structure alignment by incremental combinatorial extension (CE) of the optimal path.
Protein Eng.
11
:
739
747
.
48
Ofran
Y.
,
Schlessinger
A.
,
Rost
B.
.
2008
.
Automated identification of complementarity determining regions (CDRs) reveals peculiar characteristics of CDRs and B cell epitopes.
J. Immunol.
181
:
6230
6235
.
49
Giudicelli
V.
,
Chaume
D.
,
Lefranc
M. P.
.
2005
.
IMGT/GENE-DB: a comprehensive database for human and mouse immunoglobulin and T cell receptor genes.
Nucleic Acids Res.
33
(
Database issue
):
D256
D261
.
50
Thompson
J. D.
,
Gibson
T. J.
,
Higgins
D. G.
.
2002
.
Multiple sequence alignment using ClustalW and ClustalX.
Curr. Protoc. Bioinformatics.
Chapter 2: Unit 2.3
.
51
Clamp
M.
,
Cuff
J.
,
Searle
S. M.
,
Barton
G. J.
.
2004
.
The Jalview Java alignment editor.
Bioinformatics
20
:
426
427
.
52
Waterhouse
A. M.
,
Procter
J. B.
,
Martin
D. M.
,
Clamp
M.
,
Barton
G. J.
.
2009
.
Jalview Version 2—a multiple sequence alignment editor and analysis workbench.
Bioinformatics
25
:
1189
1191
.
53
Zhao
L.
,
Wong
L.
,
Li
J. Y.
.
2011
.
Antibody-specified B-cell epitope prediction in line with the principle of context-awareness.
IEEE/ACM Trans. Comput. Biol. Bioinformatics
8
:
1483
1494
.
54
Kuroda
D.
,
Shirai
H.
,
Kobori
M.
,
Nakamura
H.
.
2008
.
Structural classification of CDR-H3 revisited: a lesson in antibody modeling.
Proteins
73
:
608
620
.
55
Clark
L. A.
,
Ganesan
S.
,
Papp
S.
,
van Vlijmen
H. W.
.
2006
.
Trends in antibody sequence changes during the somatic hypermutation process.
J. Immunol.
177
:
333
340
.
56
Almagro
J. C.
,
Beavers
M. P.
,
Hernandez-Guzman
F.
,
Maier
J.
,
Shaulsky
J.
,
Butenhof
K.
,
Labute
P.
,
Thorsteinson
N.
,
Kelly
K.
,
Teplyakov
A.
, et al
.
2011
.
Antibody modeling assessment
.
Proteins
79
:
3050
3066
.
57
Colman
P. M.
1988
.
Structure of antibody-antigen complexes: implications for immune recognition.
Adv. Immunol.
43
:
99
132
.
58
Colman
P. M.
,
Laver
W. G.
,
Varghese
J. N.
,
Baker
A. T.
,
Tulloch
P. A.
,
Air
G. M.
,
Webster
R. G.
.
1987
.
Three-dimensional structure of a complex of antibody with influenza virus neuraminidase.
Nature
326
:
358
363
.
59
Sotriffer
C. A.
,
Liedl
K. R.
,
Linthicum
D. S.
,
Rode
B. M.
,
Varga
J. M.
.
1998
.
Ligand-induced domain movement in an antibody Fab: molecular dynamics studies confirm the unique domain movement observed experimentally for Fab NC6.8 upon complexation and reveal its segmental flexibility.
J. Mol. Biol.
278
:
301
306
.
60
Patil
A.
,
Kinoshita
K.
,
Nakamura
H.
.
2010
.
Domain distribution and intrinsic disorder in hubs in the human protein-protein interaction network.
Protein Sci.
19
:
1461
1468
.
61
Kabat
E. A.
,
Wu
T. T.
,
Perry
H. M.
,
Gottesman
K. S.
,
Foelter
C.
.
1991
.
Proteins of Immunological Interest.
National Institutes of Health
,
Bethesda, MD
.
62
Vidarte
L.
,
Pastor
C.
,
Mas
S.
,
Blazquez
A. B.
,
de los Rios
V.
,
Guerrero
R.
,
Vivanco
F.
.
2001
.
Serine 132 is the C3 covalent attachment point on the CH1 domain of human IgG1
.
J. Biol. Chem.
276
:
38217
38223
.
63
Albar
J. P.
,
Juarez
C.
,
Vivanco-Martínez
F.
,
Bragado
R.
,
Ortíz
F.
.
1981
.
Structural requirements of rabbit IgG F(ab′)2 fragment for activation of the complement system through the alternative pathway—I. Disulfide bonds.
Mol. Immunol.
18
:
925
934
.
64
Wukovitz
S. W.
,
Yeates
T. O.
.
1995
.
Why protein crystals favour some space-groups over others.
Nat. Struct. Biol.
2
:
1062
1067
.

The authors have no financial conflicts of interest.