June 21, 2006 12:7 WSPC/185-JBCB 00187

Journal of Bioinformatics and Computational BiologyVol. 4, No. 2 (2006) 415–424c© Imperial College Press

ANTIBODY CDR H3 MODELING RULES:EXTENSION FOR THE CASE OF ABSENCE OF

ARG H94 AND ASP H101

OLEG V. KOLIASNIKOV

Chemistry Department Moscow State UniversityLenGory, Moscow, Russia, 119992

olkol@enz.chem.msu.ru

MIROSLAV O. KIRAL

Kolmogorov Advanced Education and Science CenterMoscow State University

LenGory, Moscow, Russia, 119992kiralmir@mail.ru

VITALY G. GRIGORENKO∗ and ALEXEY M. EGOROV†

Chemistry Department, Moscow State UniversityLenGory, Moscow, Russia, 119992

∗vitaly@immunotek.ru†aegorov@enz.chem.msu.ru

Received 30 September 2005Revised 6 January 2006Accepted 6 January 2006

The third complementary determining region of the immunoglobulin heavy chain(CDR H3) is one of the more difficult structures to model due to genetic reasons.

However, the conformation of proximal to β-framework (“torso”) part of the CDR H3is very predictable. Current “CDR’s canonical classes” theory is based on identifyingthe key positions, H94 and H101. We can determine the CDR H3 “torso” structureif arginine or lysine is present in the H94 position and/or aspartic acid in the H101position. We target the case characterized by the absence of key residues in both theH94 and H101 positions. There has not been discussion on this case in the literature.51 CDR H3 structures of this nature are analyzed and we established new sequence-structure rules. These rules contribute to more accurate modeling of the antibody’sstructure.

Keywords: Antibody structure; CDR H3; sequence-structure correlations.

1. Introduction

Antibodies represent a protein molecule class, responsible for antigen recognitionin vertebrate organism. The structural organization of an antibody is practically

415

June 21, 2006 12:7 WSPC/185-JBCB 00187

416 O. V. Koliasnikov et al.

invariant from fishes to mammals. Common Y-like immunoglobulin quaternarystructure has hetero-tetramer constitution and consists of two heavy (∼ 500-600aa)and two light (∼ 230 aa) protein chains. The N-terminal domains of these chainsform the antigen-binding site. Due to a priori unpredictability of antigen for immunesystem, sequences of the N-terminal domains have wide diversity.

From the structural point of view, the variable parts of an antibody consist ofconserved β-framework formed by both chains and six relatively short but highlyvariable complementary determining regions (CDRs). The three regions, namelyCDRs L1, L2 and L3, originate from a light chain, while the remaindering regions(CDRs H1, H2 and H3) are segments of a heavy chain (Fig. 1).

The number of hypervariable positions takes up about 15–20% of the totalnumber of amino acid residues in the variable domains. Due to genetic reasons,CDR L3 and, especially the CDR H3, have maximum variation at the sequenceand structural levels. CDR H3 plays a major role in complete structure of antigenbinding center.

The unified structure organization of the antibodies and their wide range of prac-tical applications make structure prediction a very attractive and promising task.Whereas the sequence diversity and the resulting antigen-binding surface topologyare practically unrestricted, there are only several backbone conformations in theCDRs. Moreover, conformations of the backbone of at least five out of six CDRs canbe predicted from their sequence on the bases of simple correlation rules.1 At thispresent time, the main problem in antibody spatial structure prediction is CDR H3conformation.

The CDR H3 falls in the region of V-D-J, joined in the assembly of immunoglob-ulin heavy chain gene and is subdivided into two regions, proximal to the framework(“torso”) and the distal (“apex”). The CDR H3 “torso” is determined by the VH

Fig. 1. General view of an antibody binding site.

June 21, 2006 12:7 WSPC/185-JBCB 00187

Antibody CDR H3 Modeling Rules 417

Fig. 2. The “bulged” (left, PDB file 1dlf) and the “non-bulged” (right, PDB file 1f8t) conformationsof “torso” CDR H3. Conserved Trp H103 sidechain is shown as reference point.

and JH genes. Its conformation can be predicted more accurately than the “apex”conformation.

The “torso” follows the rules governing sequence-conformation correlation likethe other CDRs.2,3 The rules are based on the presence or absence of Arg or Aspin positions H94 and H101, respectively. Depending on the sequence, the “torso”conformation can be referred as “bulged” (i.e. formed β-bulge before conserved TrpH103) or “non-bulged” (i.e. formed extended β-hairpin) (Fig. 2).4

Both “torso” conformations are well characterized in previous reports.3,4 Unfor-tunately, the case for the absence of both residues in these positions lies beyondthe scope of this theory. Positions of H94 are coded by the VH, and H101 in theJH genes respectively. This allows us to estimate that the antibodies are 1%–2% ofthe total amount considered from the hermline gene repertoire.

The present work is devoted to filling this gap. We have compared the deter-mined antibody’s structure related to this class and formulated new rules in addi-tional to the current theory sequence-structure correlations for prediction of “torso”conformation.

2. Materials and Methods

We have analyzed 51 structures from the PDB representing 24 antibodies in ligand-complexed and free forms. Target antibody’s structure sampling was performedwith Andrew Martin’s Summary of Antibody Crystal Structures database avail-able at http://www.bioinf.org.uk/.5 Antibody atomic coordinates were obtained

June 21, 2006 12:7 WSPC/185-JBCB 00187

418 O. V. Koliasnikov et al.

from the PDB (www.rcsb.org).6 The Swiss PDB Viewer 3.7 program7 was used forvisualization. Illustrations were made with the InsightII.95 software (BioSym Inc.,San-Diego, CA, USA) on a SGI Indy R4400 workstation.

Immunoglobulin sequences were numbered according to Kabat.8 CDR H3 seg-ments were renumbered from 1 to n according to Shirai,2 that corresponds to residuerange from H95 to H102 in Kabat8 denotation.

3. Results and Discussion

As mentioned above, the primary structures of target antibodies contain neitherArg nor Asp in positions H94 (0th) and H101 ((n − 1)th), respectively. Currenttheory3 predicts “torso-bulged” (or “kinked”) conformation in all these cases (oromits this case as it is too rare4). Significant part of the selected structures has beensolved after the above theory publication. Both conformations of CDR H3 “torso”were found to contradict the theory. We pointed out this discrepancy earlier,9 butonly at present time, can general consideration be made possible due to enoughexperimental data available in the PDB database. It should be mentioned that,from our observations in all the cases, the forming of antibody-ligand complex hasno influence on the “torso” conformation assignment. Thus the only structures withdifferent primary sequences are listed in Table 1.

We have formalized “torso” structure assignment (Fig. 3). The graph is createdin Shirai coordinates.2 Dihedral angle formed by four Cα atoms at the (n − 2)th,(n − 1)th, (n)th, and (n + 1)th residues calculated and plotted on abscissa axis.Distance between backbone oxygen of (n − 2)th residue and ε-nitrogen of (n + 1)thTrp plotted on ordinate axis (Fig. 3).

From the graphs, we can observe two distinct groups of structures. Structureswith “torso-bulged” have strongly restricted O–N distance (determined by hydrogenbond) and dihedral angle (determined by β-bulge). Extended CDR H3 is charac-terized by fixed maximal O–N distance and diverse dihedral angle value. IncreasedO–N distance in 43c9 structure appears to be due to the sidechain turn in the(n + 1)th Trp with direction of ε-nitrogen of Trp into globule. Analogously, in the1ohq structure, the Trp sidechain is turned towards the domain hydrophobic core,and is further from the (n − 2)th Leu. In structures 1emt and 1h3p CDR H3 areapparently too short and their backbones adopt a slightly curved conformation.

For the selected structures, the most frequently present residue in the 0th posi-tion is Ser and in the (n − 1)th position is Ala. Its appearance in the key posi-tions does not correlate with the “torso” structure. At the same time, the 1st and(n − 2)th positions are considered to be important. Following this supposition, wefound the new sequence-structure rules describing the correlation.

The most frequently present residue in the (n − 2)th position is Phe. Amongthe 11 structures with Phe in the (n − 2)th position, eight of them have CDR H3“torso” present in “torso-bulged” conformation (PDB files 1ind, 1v7m, 1sm3, 1dlf,1f11, 1q72, 1rz7, 12e8). We observed that for all of the eight structures, interaction

June 21, 2006 12:7 WSPC/185-JBCB 00187

Antibody CDR H3 Modeling Rules 419

Table 1. Sorting of antibody structures in dependence on the “torso”CDR H3 conformation. Italics — positions H94 (0th) and H101 ((n− 1)th), underlined — (n − 2)th position, bold — 1st Tyr (H95).

Length PDB code CDR H3 sequence ......1 − n

‘‘torso-bulged’’

5 1ind CAS................................HRFVH-WG5 1v7m CSG................................WSFLY-WG5 1ktr CES................................QSGAY-WG5 2rcs CAS................................YYGIY-WG5 1ggb CVQ................................EGYIY-WG6 1sm3 CTG...............................VGQFAY-WG10 1dlf CTG..........................IYYHYPWFAY-WG10 1f11 CAN..........................DYGSTYGFAY-WG10 1q72 CTS..........................VPQLGRGFAY-WG10 1rz7 CAA..........................DPWELNAFNV-WG10 12e8 CAN .........................GHDYDRGRFPY-WG19 1gc1 CAG................VYEGEADEGEYDNNGFLKH-WG

‘‘torso-non-bulged’’

4 1emt CAT.................................SSAY-WG6 1f3d CAN...............................DYDGVY-WG7 1h3p CAS..............................FNWDVAY-WG7 1t66 CTS..............................YGYHGAY-WG7 1axt CKI..............................YFYSFSY-WG7 1jhk CAT..............................WGGNSAY-WG9 1f8t CAS...........................YDDYTWFTY-WG10 43c9 CVS..........................YGYGGDRFSY-WG10 1ncw CAG..........................LLWYDGGAGS-WG10 1yc7 CQI..........................QCGVRSIREY-WG11 1ohq CAS.........................ALEPLSEPLGF-WG12 1c12 CVT........................SLTWLLRRKRSY-WG

between the aromatic sidechain of the (n − 2)th Phe and the hydrophobic zone atthe VH-VL interface, occurs mainly with Phe L98. Hydrophobic residues of CDRL1 and H1 (Tyr L36, Val L37, Val H37, etc.) are often involved in this contact. Thebackbone oxygen of the (n − 2)th Phe forms hydrogen bond with ε-nitrogen of TrpH103 (Fig. 4).

The other 3 variants (PDB files 1axt, 1f8t, 43c9), are characterized by the pres-ence of Tyr in the 1st position in addition to the above mentioned (n − 2)th Pheand “non-bulged” conformation of CDR H3 “torso”. In the spatial structure, the1st Tyr sidechain displaced the (n − 2)th Phe and its hydroxyl group apparentlydistorted the hydrophobic joint region. At the same time, the CDR L3 is shiftedalong H3 and the Phe L98 appears in the hydrophobic zone around conserved TrpH103 (Fig. 5). We have considered the formation of “torso-non-bulged” conforma-tion in this special case in our previous work.9 For anti-2, 4-D antibody E2/B5,10 theassumption about “torso-non-bulged” conformation of CDR H3 allows us to inter-pret experimental immunochemical data on the basis of modeled spatial structure

June 21, 2006 12:7 WSPC/185-JBCB 00187

420 O. V. Koliasnikov et al.

0 50 100 150 200 2500,0

0,2

0,4

0,6

0,8

1,0

1,2

1,4

1h3p1emt

1ohq

43c9

"torso-non-bulged"

"torso-bulged"

Dis

tanc

e O

-N, n

m

Dihedral angle, grad

Fig. 3. Distribution of dihedral angles and inter-atomic distances in CDR H3. Dihedral anglesare measured between four Cα atoms of (n − 2)th, (n − 1)th, (n)th and (n + 1)th residues.The inter-atomic distance is measured between the backbone oxygen of (n − 2)th residue andε-nitrogen of (n + 1)th Trp.

Fig. 4. Selected sidechains conformation view in the case of “torso-bulged” (PDB file 1dlf).

and to plan for successful mutagenesis experiment for rational design of antibodyspecificity (in preparation).

For the last considered variant with the presence of Gly in the (n − 2)th or(n − 3)th positions, it is found to correlate with the CDR size. In the case of veryshort 5-residue loop (PDB files 2rcs, 1ktr, 1ggb), the CDR H3 backbone adopts

June 21, 2006 12:7 WSPC/185-JBCB 00187

Antibody CDR H3 Modeling Rules 421

Fig. 5. Selected sidechains conformation view in “torso-non-bulged” case (PDB file 1f8t).

Fig. 6. S-like CDR H3 conformation view in case of 5-residue size and the Gly in the (n − 2)thposition (PDB file 2rcs).

June 21, 2006 12:7 WSPC/185-JBCB 00187

422 O. V. Koliasnikov et al.

the S-like conformation, which consists of the β-bulge followed by the β-turn withGly residue between them (Fig. 6). This results in the observed “torso-bulged” con-formation. The additional residues (PDB files 1t66, 1f3d) has seemingly disturbedthis harmony and the “torso” straightens out. This observation is not certain dueto deficient experimental data.

The most part of the other structures (7 of 8) has “torso-non-bulged” confor-mation except the case (PDB file 1gc1) with very large 19-residue CDR H3.

4. Conclusion

For the first time, we supplemented the structure-conformation correlations forCDR H3 “torso” conformation prediction in the case of absence of Arg or Asp inpositions H94 and H101, respectively. The followings are concluded for the targetedcase:

(1) When Phe is present in the (n − 2)th position and Tyr is absent in the 1stposition, Phe interacts with the hydrophobic zone on VH-VL interface, and“torso-bulged” structure is formed.

(2) When Phe is present in the (n − 2)th position and Tyr is present in the 1stposition, the 1st Tyr sidechain distorts the hydrophobic contact, and “torso-non-bulged” structure is formed.

(3) Presumably, when CDR H3 has 5-residue size and Gly is present in the(n − 2)th or (n − 3)th position, the “torso-bulged” structure is formed. Inmost of other cases, “torso-non-bulged” conformation is formed.

These rules agree with 23 out of the 24 different CDR H3 structures. Thus, thecorrelations found contribute to more accurate antibody structure modeling.

Acknowledgments

We would like to acknowledge Dr. Igor Ouporov for useful discussion of this workduring preparation to publishing and the help of Swee Seong Wong in improvingthe English of the manuscript.

References

1. Chothia C, Lesk AM, Tramontano A et al., Conformations of immunoglobulin hyper-variable region, Nature 342:877–883, 1989.

2. Shirai H, Kidera A, Nakamura H, Structural classification of CDR-H3 in antibodies,FEBS Lett 399:1–8, 1996.

3. Shirai H, Kidera A, Nakamura H, H3-rules: identification of CDR-H3 structures inantibodies, FEBS Lett 455:188–197, 1999.

4. Morea V, Tramontano A, Rustici M et al., Conformations of the third hypervariableregion in the VH domain of immunoglobulins, J Mol Biol 275:269–294, 1998.

5. Martin AC, Thornton JM, Structural families in loops of homologous proteins:automatic classification, modelling and application to antibodies, J Mol Biol263:800–815, 1996.

June 21, 2006 12:7 WSPC/185-JBCB 00187

Antibody CDR H3 Modeling Rules 423

6. Berman HM, Westbrook J, Feng Z et al., The protein data bank, Nucleic Acids Res28:235–242, 2000.

7. Guex N, Peitsch MC, SWISS-MODEL and the Swiss-Pdb Viewer: An environmentfor comparative protein modeling, Electrophoresis 18:2714–2723, 1997.

8. Kabat EA, Wu TT, Perry HM et al., Sequences of proteins of immunological interestBethesda, MD: NIH, USA, 1991.

9. Koliasnikov OV, Grigorenko VG, Egorov AM, Analysis of the model binding site ofanti-2,4-dichlorophenoxyacetic acid antibodies, Biomed Khim 49(3):238–249, 2003.

10. Franek M, Kolar V, Granatova M, Nevorankova Z, Monoclonal ELISA for2,4-dichlorophenoxyacetic acid: characterization of antibodies and assay optimization,J Agric Food Chem 42:1369–1374, 1994.

Oleg V. Koliasnikov — born in 1976 in Volgograd, Russia.1998 — graduated from Moscow State University. Ph.D. student(1999–2002) in Enzymology division of Chemistry Departmentunder the supervision of Professor Alexey M. Egorov. Ph.D. workon antibody structure modeling and protein engineering. Cur-rently a research scientist at the Moscow State University, ChemDept., laboratory of enzyme engineering. Also he is a teacher inKolmogorov Advanced Education and Science Center.

Miroslav O. Kiral — born in 1989 in Schoekino, Tula region(Russia).

Since 1994 he has been studying in Kolmogorov AdvancedEducation and Science Center of Moscow State University.Presently, he is a student of Chemical Faculty of M.V.Lomonosov Moscow State University.

Vitaly G. Grigorenko — born in 1968, Moscow, Russia.1991 — graduated from Moscow Institute of Physics and Tech-nology, division of Physico-chemical Biology and Biotechnol-ogy; Ph.D. in biotechnology at M.V. Lomonosov Moscow StateUniversity. (2001) (Thesis — Recombinant horseradish peroxi-dase: production, properties and application). Currently a seniorresearch scientist at Moscow State University, Chem Dept.,laboratory of enzyme engineering, group leader.

June 21, 2006 12:7 WSPC/185-JBCB 00187

424 O. V. Koliasnikov et al.

Alexey M. Egorov — born 1943, Moscow, Russia. Education:Student of Biology-Soil Department, Lomonosov, Moscow StateUniversity, Russia, 1961–1966; Ph.D., 1971; Thesis: “The elu-cidation of human immuno-globulins quaternary structure”,Moscow State University, Russia; Doctor Biological ScienceDegree, 1985; Thesis: “The structure and kinetic mechanismaction of NAD-dependent C1-compounds dehydrogenases andbased on them the creation of cofactor regeneration system”

Moscow State University, Russia; Professor 1993, (Biochemistry), Moscow StateUniversity, Russia; Correspondent-member of Russian Academy of Medical Science1995, Russia Academician of Russian Academy of Medical Science 2000, Russia.