S

Cm

AYa

b

a

ARRAA

KAAACCMNPPX

1

n(Nesiacs2e

(

F

0d

Molecular Immunology 46 (2009) 3466–3475

Contents lists available at ScienceDirect

Molecular Immunology

journa l homepage: www.e lsev ier .com/ locate /mol imm

hort communication

rystal structure of an anti-ganglioside antibody, and modelling of the functionalimicry of its NeuGc-GM3 antigen by an anti-idiotypic antibody

riel Talaveraa,b,1, Agneta Erikssonb,1, Mats Ökvistb,2, Alejandro López-Requenaa,uniel Fernández-Marreroa, Rolando Péreza, Ernesto Morenoa,∗∗, Ute Krengelb,∗

Center of Molecular Immunology, P.O. Box 16040, Havana 11600, CubaDept. of Chemistry, University of Oslo, P.O. Box 1033, Blindern, NO-0315 Oslo, Norway

r t i c l e i n f o

rticle history:eceived 27 May 2009eceived in revised form 24 July 2009ccepted 28 July 2009vailable online 12 September 2009

eywords:ntigen mimicrynti-idiotypic antibody

a b s t r a c t

N-Glycolylated (NeuGc) gangliosides are tumor-specific antigens and as such represent attractive targetsfor cancer immunotherapy. The chimeric antibody chP3 selectively recognizes a broad variety of NeuGcgangliosides, showing no cross-reactivity to the highly similar N-acetylated (NeuAc) gangliosides thatare common cellular antigens in humans. Here, we report the crystal structure of the chP3 Fab and itscomputer-docking model with the trisaccharide NeuGc�3Gal�4Glc�, which represents the carbohydratemoiety of the tumor-antigen NeuGc-GM3. The interaction involves only the heavy chain of the chP3 anti-body. The modelled complex is consistent with all available experimental data and shows good surfacecomplementarity. The negatively charged sialic acid residue NeuGc is buried in a pocket flanked by two

nti-tumor antibodiesancer vaccineomputer dockingolecular modellingeuGc-GM3 gangliosiderotein–carbohydrate interactionsrotein–protein complex

arginine residues, VH Arg31 and VH Arg100A. We have further investigated the interaction of chP3 withits anti-idiotypic antibody, 1E10 (also known as Racotumomab), currently in clinical trials as a cancervaccine. While many of the chP3 residues predicted to interact with the NeuGc ganglioside also featureprominently in the modelled complex of chP3 and 1E10, we do not observe structural mimicry. Rather,we suspect that the anti-idiotype 1E10 may serve as an imprint of the structural characteristics of thechP3 idiotype and, consequently, give rise to antibodies with P3-like properties upon immunization.

© 2009 Elsevier Ltd. All rights reserved.

-ray crystal structure

. Introduction

P3 is a monoclonal antibody (mAb) that was generated by immu-izing mice with the ganglioside N-glycolyl-GM3 (NeuGc-GM3)Vázquez et al., 1995). P3 is able to recognize a broad variety of-glycolyl-containing gangliosides as well as sulfatides (Vázquezt al., 1995; Moreno et al., 1998). NeuGc gangliosides can be con-idered to be tumor-specific antigens, because their expressionn humans is practically limited to tumor tissues (Marquina etl., 1996; Malykh et al., 2001; Kannagi et al., 2008). However,

arbohydrates usually do not elicit memory immune responsesince they are T-cell independent antigens (Lesinski and Westerink,001). One of the alternatives for active cancer therapy is the gen-ration of idiotypic determinants that “mimic” the carbohydrate

∗ Corresponding author. Tel.: +47 22855461; fax: +47 22855441.∗∗ Corresponding author.

E-mail addresses: emoreno@cim.sld.cu (E. Moreno), ute.krengel@kjemi.uio.noU. Krengel).

1 These authors have contributed equally to this paper.2 Present address: ESRF, 6 rue Jules Horowitz, BP 220, 38043 Grenoble Cedex 9,

rance.

161-5890/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.oi:10.1016/j.molimm.2009.07.032

epitopes. Several anti-idiotypic antibodies specific for P3 have beendescribed (Vázquez et al., 1998; López-Requena et al., 2007b). Oneof them, named 1E10 (Vázquez et al., 1998) (also known as Racotu-momab), has been used in clinical trials in patients with melanoma(Alfonso et al., 2002), breast cancer (Díaz et al., 2003; Guthmann etal., 2006), small cell lung cancer (Neninger et al., 2007) and non-small cell lung cancer (Hernández et al., 2008), with encouragingresults. Most of these patients developed anti-NeuGc-GM3 anti-bodies upon immunization with a vaccine consisting of 1E10 mAbadjuvated in alum.

A human IgG1 chimeric version of P3 (chP3) was constructed,maintaining the recognition and immunogenic properties that havebeen described for the original murine IgM P3 (López-Requena etal., 2003). The chP3 antibody specifically reacted with NeuGc-GM3and NeuGc-GM2 gangliosides, while it did not bind to their respec-tive acetylated variants, which are common cellular antigens inhumans. Such discrimination is remarkable given the minor differ-

ence between the two molecules, consisting of one added oxygenatom for NeuGc compared to NeuAc.

Our earlier work on immunogenetic analysis (Perez et al., 2001),epitope mapping studies with phagotopes (Perez et al., 2001;López-Requena et al., 2007a) as well as structural modelling (López-

Immun

Ri(eaVo1f2recRtrbt2

fa2ba2eere((a

asgmiopt“

2

2

(dcmi(tFs

2

Me

A. Talavera et al. / Molecular

equena et al., 2007a) of the P3 and 1E10 mAbs identified severalmportant charged residues at their heavy chain variable domainVH) complementarity determining regions (CDRs), which mayither complement or “mimic” the negative charge of the sialiccid residue NeuGc. P3 has surface-exposed arginine residues atH CDR1 and CDR3 that could be important for the recognitionf the ganglioside and also for interaction with the VH CDRs fromE10, which are enriched in acidic residues. This latter feature wasound to be shared by other anti-idiotypic mAbs of P3 (Perez et al.,001; López-Requena et al., 2007b). The importance of the arginineesidues for P3-ligand interaction was demonstrated by mutagen-sis studies. The introduction of serine residues at those positionsaused the loss of the reactivity towards the ganglioside (López-equena et al., 2007a). Surprisingly, in contrast to the consensushat VH CDR3 rules most of the antibody interactions, it was theeplacement of the arginine residue at VH CDR1 that abolished theinding to all of the P3-specific anti-idiotypic mAbs, while muta-ions at VH CDR3 had little or no effect (López-Requena et al.,007a,b).

The presence of arginine motifs at VH CDR3 has been reportedor a number of autoantibodies, particularly anti-nuclear (Chen etl., 2002) and anti-phospholipid (Cocca et al., 2001; Giles et al.,003) antibodies. The importance of these residues for the anti-ody specificity has been demonstrated (Guth et al., 2003; Giles etl., 2005). Another anti-NeuGc-GM3 mAb, named 14F7 (Carr et al.,000), also exhibits a VH CDR3 enriched in arginines (Rodríguezt al., 2007). The crystal structure of the 14F7 Fab and its mod-lled complex with the antigen point to the involvement of theseesidues in the molecular recognition of the ganglioside (Krengelt al., 2004). An anti-idiotypic mAb specific for 14F7, named 4G9Rodríguez et al., 2003), displays acidic amino acids at its VH CDR3Rodríguez et al., 2007), which may interact with the basic residuest the 14F7 combining site, like in the P3–1E10 case.

Here, we report the crystal structure of the chP3 Fab fragmentt 1.8 Å resolution. Using computer simulations, we predict thetructure of the chP3 complex with its ligand, the NeuGc-GM3anglioside. Furthermore, the structure of the 1E10 variable frag-ent (Fv) was modelled and used to predict the structure of the

diotype–anti-idiotype complex by protein–protein docking. Thebtained models are supported by a wealth of experimental data,art of them reported in this work. Based on the comparison of thewo models, we propose a possible explanation for the so-calledantigen mimicry”.

. Materials and methods

.1. Cells

NS0 transfectomas expressing chP3 mAb, the P3 light chainLópez-Requena et al., 2003) and its mutants, the hybridomas pro-ucing the murine 1E10 mAb (Vázquez et al., 1995) and the murineell line L1210 (purchased from ATCC) were cultured in Dulbecco’sodified Eagles medium (DMEM) supplemented with 10% heat

nactivated fetal calf serum (FCS), antibiotic mixtures of penicillin100 U/ml), streptomycin (100 �g/ml) and 2 mM l-glutamine. Forhe selection of whole antibody-producing transfectomas, DMEM-12 containing 10% FCS and histidinol at 10 mM was used aselective medium.

.2. Gangliosides and monoclonal antibodies

The NeuGc-GM3 ganglioside was produced at the Center ofolecular Immunology (Havana, Cuba). It was isolated from horse

rythrocytes as previously described (Stults et al., 1989).

ology 46 (2009) 3466–3475 3467

1E10 mAb (Vázquez et al., 1998), chP3 (López-Requena etal., 2003) and VH CDR3 mutants of chP3 (human IgG1, kappa)were purified by protein A affinity chromatography (Pharmacia,Uppsala, Sweden) and analyzed by SDS PAGE under reducingconditions. Murine antibodies were purified from ascitic fluid,whereas chimeric antibodies were purified from transfectoma cul-ture supernatants. The specificity of the purified antibodies wasconfirmed by enzyme-linked immunosorbent assay (ELISA) and theprotein concentration was estimated by optical density at 280 nm.

2.3. Fab fragment preparation and purification

Fab fragments of chP3 were prepared from whole mAb by papaindigestion. chP3 mAb (2–3 mg, concentration ranging from 1 to2 mg/ml, in 20 mM phosphate buffer pH 7.0) was digested withpapain in the presence of l-cysteine and EDTA for 4 h at 37 ◦C. Theratio of mAb:papain was 100:1 and the final concentrations of l-cysteine and EDTA were 10 mM each. The reaction was stopped byadding iodoacetamide to a final concentration of 20 mM. The Fabfragments were purified on a protein A column (HiTrapTM ProteinA HP Column 5 ml, GE Health Care), which binds Fc fragments andundigested mAb while the Fab fragments elute in the flow-through.The buffer of the purified Fab was thereafter changed to 25 mM TrispH 7.0 and the protein was concentrated to 10–12 mg/ml.

2.4. Crystallization and data collection

Crystals of chP3 Fab were obtained from different conditionsand tested for X-ray diffraction. All crystallization trials were per-formed with the vapor diffusion hanging drop technique at 20 ◦C.Final data were collected on a crystal obtained from crystallizationconditions containing 30% polyethylene glycol (PEG) 10 K, 0.1 Mcacodylate pH 6.5 and 0.2 M ammonium sulfate, which had grownto a size of 0.2 mm × 0.1 mm × 0.1 mm over 7.5 months. 15% glyc-erol was used as a cryo-protectant. The data were collected intwo passes, to 1.75 and 3.0 Å resolution, respectively, at beamlineID 14-1 at the European Synchrotron Radiation Facility (ESRF) inGrenoble, France. Data were processed with iMosflm 0.6.1 (builton the latest Mosflm 7.0.3) (Collaborative Computational Project,Number 4 (CCP4), 1994; Leslie, 2006) and the intensities fromthe two data sets were thereafter scaled and merged with SCALA(CCP4, 1994; Evans, 2006). The program PointLess (CCP4, 1994;Evans, 2006; Leslie, 1992) was used to determine the correct spacegroup (I422). The solvent content was estimated to be 62% based onthe Matthew’s coefficient (Matthews, 1968; Kantardjieff and Rupp,2003) (VM = 3.2 Å3/Da). Data collection statistics are summarized inTable 1.

2.5. Structure determination and refinement

The structure of the chP3 Fab was solved by molecular replace-ment with the program PHASER (McCoy, 2007), using the structureof the CAMPATH-1H Fab (PDB code: 1ce1; James et al., 1999) as asearch model. The program PHENIX AutoBuild (Zwart et al., 2008)was used for model building and to obtain improved maps. Thesemaps were manually inspected and the autobuilt segments wererearranged to form a compact molecule and completed for themissing segments. PHENIX refine (Adams et al., 2002) was usedto perform simulated annealing and to add solvent molecules tothe model, followed by automated model building with ARP/wARP

(Perrakis et al., 1999). Finally, the structure was manually rebuiltusing the program COOT (Crystallographic Object-Oriented Toolkit)(Emsley and Cowtan, 2004). The model was refined with REFMAC5(Murshudov et al., 1997). Refinement statistics are summarized inTable 1.

3468 A. Talavera et al. / Molecular Immu

Table 1Data collection and refinement statistics.

Space group I422a = b (Å) 153.5c (Å) 111.5

Data collection statisticsResolution range (Å) 55.7–1.75 (1.795–1.75)No. of unique reflections 63,026 (9658)Multiplicity 8.8 (7.3)Rmergea (%) 6.6 (51.5)Mean I/� (I) 19.2 (4.2)Completeness (%) 99.4 (99.9)

Refinement statisticsRwork

b (%) 19.4Rfree (%) 23.4r.s.c.cc 0.95r.m.s.dd, bonds (Å) 0.022r.m.s.d, angles (◦) 1.96Average B-factor (Å2) 26.4Wilson B-factor (Å2) 22.8

Final modelNo. of Fabs in asymmetric unit 1No. of amino acid residues 429No. of water molecules 422No. of other molecules 5 (SO4

2−)

Ramachandran statisticse

Most favored (%) 89.9Allowed (%) 10.3Generously allowed (%) 0.5Disallowed (%) 0.3

Numbers in parentheses refer to high resolution shell.a Rmerge = ˙h˙i|Ihi − 〈Ih〉|/˙h˙i〈Ih〉 where 〈Ih〉 is the average intensity over

symmetry-related measurements.b Rfactor = ˙||FO| − |FC||/˙|FO| where FO and FC are the observed and calculated

structure factors, respectively.c

1

2

s112

Lfcs

r.s.c.c.: real space correlation coefficient.d r.m.s.d.: root mean square deviation from ideal geometry (Engh and Huber,

991).e According to PROCHECK (Laskowski et al., 1993).

.6. Construction and expression of VH CDR3 mutants of chP3

The arginine residue at position 100A of chP3 VH CDR3 wasubstituted by a threonine residue, and tryptophan at position00E by a tyrosine residue, by overlapping PCR (Kammann et al.,989), using the VHP3-pAH4604 vector (López-Requena et al.,003) as a template and the oligonucleotides:

The sequences of the two first oligonucleotides were taken fromópez-Requena et al. (2003). Restriction sites for EcoRV and NheI,or ‘leader P3’ and ‘JH P3’, respectively, are underlined. Mutatedodons are in italics. Numbers above each codon follow the Kabatystem. The oligonucleotide ‘leader P3’ was combined with VH

nology 46 (2009) 3466–3475

Arg100A → Thr-3′ or VH Trp100E → Tyr-3′ and ‘JH P3’ with VHArg100A → Thr-5′ or VH Trp100E → Tyr-5′ for mutagenic poly-merase chain reaction (PCR). PCR “together” were performed witholigonucleotides ‘leader P3’ and ‘JH P3’.

The purified PCR products were cloned into the pMOS-Blue vec-tor (Amersham Pharmacia, UK), following the instructions of theprovider. Clones were sequenced by the dideoxy method (Sangeret al., 1977) using the T7 DNA Pol kit (Pharmacia). After checkingthe sequences, the DNA fragments were digested with EcoRV/NheIand cloned into the pAH4604 vector, containing the human IgG1constant region (IGHG1*01 allele). The pAH4604 vector for heavychain expression (histidinol selectable marker) has been describedin detail (Coloma et al., 1992) and was kindly supplied by Dr. SherrieL. Morrison, Department of Microbiology and Molecular Genetics,UCLA, USA.

The NS0 transfectoma cells expressing the chP3 light chain(López-Requena et al., 2003) were transfected by electroporationwith the pAH4604 expression vector containing the chP3 heavychain with mutated VH CDR3, and chimeric antibody expressionwas determined by ELISA, as described elsewhere (López-Requenaet al., 2007a).

2.7. Antibody binding assays

Binding of chP3 VH CDR3 mutant antibodies to gangliosides wasdetermined by ELISA, as previously described (López-Requena etal., 2007a). Similarly, binding of chP3 VH CDR3 mutant antibodiesto anti-idiotypic mAbs was determined by ELISA, as described forthe previous VH CDR3 Arg → Ser mutants (López-Requena et al.,2007a,b). Three samples were tested in each experiment and thecoefficient of variation was <10% for all assays. Background valuesof absorbance were less than 0.1.

Non-competitive ELISA for affinity determination was per-formed as previously described (Beatty et al., 1987). Briefly, serialdilutions of purified chP3 were incubated on 96-well polystyrenemicrotiter plates (High binding, Costar, Cambridge, MA) coatedwith different 1E10 mAb concentrations. Bound chP3 was detectedwith an alkaline phosphatase-conjugated goat anti-human IgG(gamma-chain-specific) antibody (Sigma, Saint Louis, MO). Curveswere obtained for each antigen concentration, and equilibriumconstants were calculated from them as established (Beatty et al.,1987).

Staining of L1210 cells coincubated with different concen-trations of chP3 and biotin-conjugated ch14F7 mAb (10 �g/ml),followed by addition of FITC-conjugated streptavidin (BD Pharmin-gen), was determined using a FACScan instrument (Becton

Immun

Dt

2l

ba(grtraiceab

2

paordCwwcp(

aatwLAEawoa

2

Rcufu1mtoae2a

A. Talavera et al. / Molecular

ickinson, Franklin Lakes, NJ). The WinMDI 2.8 program was usedo analyze a total of 5000 cells acquired on every FACS assay.

.8. Modelling the 1E10 variable domain and the long VH CDR3oops of P3 and 1E10

A computer model of the 1E10 mAb variable domain wasuilt from the light chain variable domain (VL) and VH aminocid sequences using the Web Antibody Modelling (WAM) serverWhitelegg and Rees, 2000) (http://antibody.bath.ac.uk/). Since theeometry of a long VH CDR3 in a modelled Fv structure is noteliable, we used the program Modeller (Sali and Blundell, 1993)o generate different conformations for the five most exposedesidues of the 1E10 VH CDR3 loop. We employed this programlso to model the tip (three residues) of chP3 VH CDR3, whichs not modelled in the crystal structure. For each antibody, 100onformations of the VH CDR3 were built. Thereafter, the differ-nt loop geometries were clustered by their backbone and C�tom coordinates, using two different RMSD cutoff values (seeelow).

.9. Docking simulations of the chP3–sugar complex

The program AutoDock 4.0 (Morris et al., 1998) was used toerform docking simulations between chP3 and the NeuGc sialiccid residue. The starting structure of NeuGc was built basedn topology files included in the Quanta2000 package (Accel-ys Inc.). Nine representative structures of the chP3 variableomain, differing from each other in the conformation of VHDR3, were used for independent docking runs. These structuresere obtained by clustering the 100 conformations generatedith the program Modeller, as described above, using an RMSD

utoff value of 1.5 Å. The ligand and receptor structures were pre-ared for the docking simulations using the AutoDock Tools suitehttp://autodock.scripps.edu/resources/adt/index html).

The flexibility of most NeuGc functional groups was taken intoccount in the docking simulations (eight torsion angles werellowed to vary). Since AutoDock allows a limited degree of recep-or flexibility, three side chains located in the putative binding siteere set as flexible: VH Arg31, Ser56 and Arg100A. 200 runs of the

amarckian genetic algorithm (Morris et al., 1998) implemented inutoDock were executed for each of the nine receptor structures.ach run was performed with an initial population of 300 individu-ls, to complete a maximum of 10,000 generations. Default valuesere used for the rest of the parameters. The resulting 1800 ligand

rientations were clustered and analyzed using home-made scriptsnd the program VMD (Humphrey et al., 1996).

.10. Docking simulations of the chP3/1E10 complex

The docking simulations were performed using the programosettaDock (version 2.0) (Gray et al., 2003). The 100 VH CDR3onformations obtained for both chP3 and 1E10 were clusteredsing a 2 Å RMSD cutoff, yielding five representative VH CDR3 con-ormations for each antibody. A total of 25 starting models weresed, combining the five different models of chP3 with those ofE10. Taking into account that both antibodies have to interact

ainly through their hypervariable loops, we used the perturba-

ion mode of the program RosettaDock to focus the explorationn this molecular interface. A total of 25,000 decoys were gener-ted (1000 for each of the 25 starting models). The results fromach run were pooled and ranked by their energy scores, and the0 top-scoring (lowest energy) decoys were selected for furthernalysis.

ology 46 (2009) 3466–3475 3469

2.11. Computational mutagenesis

For computational mutagenesis studies we used the Roset-taInterface (Kortemme and Baker, 2002) software. This programestimates the change in the binding free energy (��G) forindividual mutations or a combination of mutations, comparedto the wild-type complex. We used the previously reported cri-terion (Kortemme et al., 2004) of |�Gmut − �Gwt| > 1 kcal/mol topredict mutations able to affect (��G > 1) or improve (��G < −1)protein–protein binding, while |�Gmut − �Gwt| < 1 kcal/mol corre-lates with mutations that should not affect binding.

2.12. Antibody numbering scheme

In order to maintain consistency with previous reports on chP3and 1E10 mAb, we used here the Kabat numbering scheme for theantibody variable domains, which for chP3 introduced insertionletters in the VH sequence after positions 82 (82A to 82C), and100 (100A to 100F), while no insertions were introduced in the VLsequence. For 1E10, the insertions in the VH sequence are 52A, 82Ato 82C and 100A to 100D, while the VL sequence has no insertions.

3. Results and discussion

3.1. Crystal structure of the chP3 Fab fragment

The crystal structure of the chP3 Fab fragment was determinedat 1.8 Å resolution and refined to crystallographic R/Rfree-factors of19.4 and 23.4%, respectively, with good stereochemistry (Table 1).The structure of the Fab fragment shows the classical immunoglob-ulin fold. The electron density is very well-defined for most ofthe structure, with a real space correlation coefficient (r.s.c.c.) of95%, except for two loop regions comprising heavy chain residues99–100A (the tip of VH CDR3) and 130–133 (a loop in the CH1constant domain). For CH1 130–133, electron density is lacking,while for VH CDR3, difference density is clearly present, but notsufficiently well-defined to allow model building with confidence(Fig. 1B). The structure has therefore not been modelled for theseparts. Mutagenesis studies have identified the VH CDR3, togetherwith VH CDR1, as important part of the antigen-binding epitope(López-Requena et al., 2007a). In the crystal structure, the VH CDR1loop is largely (though not completely) solvent-exposed, whereasone of the important arginine residues in VH CDR1, Arg31, isinvolved in strong van der Waals interactions with a symmetry-related molecule in the crystal. The long VH CDR3 loop (14 aminoacid residues) protrudes from the middle of the binding region,partially separating the VH and VL moieties of the combining site(Fig. 1A and C), as observed earlier for the 14F7 antibody (1rih;Krengel et al., 2004), which has a similar specificity as chP3. Asa consequence, chP3 does not show a central binding cavity, astypically found in antibodies that bind small molecules.

3.2. Site directed mutagenesis

We tested by ELISA the effect of two single mutations in the chP3heavy chain, Arg100A → Thr and Trp100E → Tyr, on the recognitionof either the NeuGc-GM3 ganglioside or the P3 anti-idiotype 1E10(Fig. 2). Mutation of VH Trp100E → Tyr had no effect on gangliosidebinding, while the VH Arg100A → Thr substitution completely abol-ished binding to NeuGc-GM3 (Fig. 2A). The binding of 1E10 was notaffected by these two mutations (Fig. 2B), as previously reported for

two other single mutants, VH Arg98 → Ser and VH Arg100A → Ser(López-Requena et al., 2007a). While Arg100A is located at the tipof VH CDR3 (Fig. 1B), Trp100E is placed at the base of this CDR andis buried at the VH–VL interface. Modelling this substitution onthe crystal structure shows that the tyrosine side chain at position

3470 A. Talavera et al. / Molecular Immunology 46 (2009) 3466–3475

Fig. 1. Structures of the chP3 Fab and 1E10 Fv fragments. (A) Cartoon representation of the chP3 Fab crystal structure. The light chain is shown in wheat and the heavy chainin green. CDR loops are labeled H1-3 and L1-3 for VH CDR1-3 and VL CDR1-3, respectively. The missing VH CDR3 tip is represented by a dashed line. (B) An attempt was madet Fo − Fc the VHs d fromd 3 and

Vt

3

pmDafltabda

bttstct

o model VH CDR3 loop residues 99–100A into the electron density (�A-weighted 2hP3, showing the five alternative computer-generated backbone conformations ofhows highest resemblance to the conformation in the crystal structure, as inferreisplaying the five different conformations of its VH CDR3 used for docking. Figs. 1,

H 100E can play the same structural role as the indole group ofryptophan.

.3. Model of the chP3/NeuGc-GM3 complex

We decided to perform docking simulations to model the com-lex of the chP3 Fab fragment and the NeuGc-GM3 trisaccharideoiety, in parallel to ongoing efforts to crystallize this complex.ocking was performed with the program AutoDock 4.0 (Morris etl., 1998), keeping the small ligand and a few protein side chainsexible. Since protein–ligand docking algorithms are very sensitiveo the conformation of the binding site (Teodoro and Kavraki, 2003)nd, on the other hand, AutoDock does not handle backbone flexi-ility, we selected nine representative chP3 Fv structures showingifferent backbone conformations of the tip of VH CDR3 (Fig. 1C),s described in Section 2.

The docking simulations were focused on the VH side of the anti-ody’s binding region, based on two main considerations. Firstly,he VH domain contains Arg31, which has been shown to be impor-

ant for binding of NeuGc-GM3 (López-Requena et al., 2007a) and,econdly, since the antigen is relatively small and has a linear struc-ure, it cannot bind to both the VL and VH sides of the antibodyombining region simultaneously, separated from each other byhe protruding VH CDR3.

c electron density map contoured at 1�). (C) Representation of the Fv fragment ofCDR3 loop tip that were used for docking to the 1E10 antibody. The loop in yellowthe electron density shown in B. (D) Computer model of the 1E10 Fv fragment,

5 were made using PyMol (http://pymol.sourceforge.net/).

The AutoDock runs resulted in 1800 ligand conformations, withabout 40% of them showing energy scores within a narrow intervalof 3 kcal/mol above the lowest energy. Clustering of this subset ofsolutions yielded 83 different representative conformations, whichwere individually inspected to evaluate their fitness, according to aset of five experimentally based criteria: (1) the N-glycolyl group ofNeuGc, and especially its distinctive hydroxyl group, should be intight contact with the antibody, since it is this group that defines thespecificity of chP3 for NeuGc-containing glycolipids (Vázquez et al.,1995; Moreno et al., 1998); (2) the carboxyl group of the sialic acidshould interact with the protein, because chemical derivatizationof this group abolished P3 binding (Moreno et al., 1998); (3) theorientation of the sialic acid residue should allow the placement ofthe Gal� residue in one of the two conformations observed for theNeuGc� 3Gal� disaccharide (Siebert et al., 1992), without clashinginto the antibody; (4) likewise, the orientation of the NeuGc� 3Gal�disaccharide should allow connecting, at position 4 of the galactose,the GalNAc� residue present in the NeuGc-GM2 ganglioside, whichis also recognized by P3 (Vázquez et al., 1995); and (5) VH Arg31should participate in the interaction, since mutation of this residue

abolished binding of chP3 to NeuGc-GM3 (López-Requena et al.,2007a).

From the 83 representative conformations that were analyzed,only one structure, belonging to a 9-member cluster, satisfied allfive criteria. The structure of the obtained complex is shown in

A. Talavera et al. / Molecular Immun

Fig. 2. Binding properties of VH CDR3 mutants of chP3. (A) Recognition of purifiedNeuGc-GM3 ganglioside by ELISA. The plate wells were coated with 500 ng of gan-glioside in 50 �l of methanol. Antibodies were used at a concentration of 10 �g/ml.(B) Recognition of the anti-idiotypic 1E10 mAb. The plate wells were coated with1c

FcaNfacc(tegcppbbr

eg(gwwttithegca

0 �g/ml of 1E10 mAb. The recognition was measured using serial dilutions of theulture supernatants from the transfectomas producing the chimeric antibodies.

ig. 3(panels A and B). In this model of the chP3/NeuGc-GM3omplex, all the functional groups of NeuGc participate in inter-ctions with the antibody. The distinctive hydroxyl group of the-glycolyl group is buried in a small, highly hydrophilic pocket,

ormed by the backbone NH and the hydroxyl group of VH Ser33nd the hydroxyl group of VH Ser95, while the hydroxymethylarbon of the NeuGc N-glycolyl group is in contact with the sidehains of VH Val97 and VH Ala100B and with C� of VH Tyr32Fig. 3B). Up to three hydrogen bonds may be established betweenhe N-glycolyl OH and the surrounding polar groups, which mayxplain the fine distinction between NeuGc- and NeuAc-containingangliosides shown by chP3. It is worth noting that in the chP3rystal structure a water molecule was found within 2 Å from theosition occupied by the N-glycolyl OH-group in the modelled com-lex. The N-glycolyl group is further involved in two hydrogenonds – from its carbonyl oxygen and the NH group to the back-one NH of VH Ala100B and the carbonyl oxygen of VH Arg31,espectively.

The model is also consistent with binding data obtained for sev-ral chemical derivatives of NeuGc-GM3, in which the carboxylroup and the glycerol tail of the sialic acid residue were modifiedMoreno et al., 1998). The carboxylate of NeuGc is found facing theuanidinium group of VH Arg31, forming a bi-dentate salt bridgeith this residue. The glycerol tail makes hydrophobic contactsith the indole ring of VH Trp52 while being partially exposed to

he solvent, in agreement with experimental results showing thatrimming of this tail affected, but did not abolish antibody bind-ng (Moreno et al., 1998). The galactose residue is mostly exposedo the solvent, only making contact with VH Gly100. Due to the

igh flexibility of the VH Arg100A side chain, it may establish sev-ral different contacts both with the glycerol tail of NeuGc and thealactose residue, and also with the hydroxymethyl group of glu-ose. Otherwise, the glucose residue remains separated from thentibody surface.

ology 46 (2009) 3466–3475 3471

3.4. Comparison of the models of the chP3/NeuGc-GM3 and14F7/NeuGc-GM3 complexes

The interaction of chP3 with NeuGc-GM3, according to the com-puter model presented in this work, has many common featureswith the modelled complex of the 14F7 antibody with the same lig-and (Krengel et al., 2004): (1) both antibodies have long VH CDR3loops that protrude from the middle of the binding site, dividing itinto two separated clefts; (2) only the heavy chains participate inthe interaction with the antigen, which involves the three VH CDRs;(3) the two antibodies interact mainly with the sialic acid residue,while the galactose unit contributes only with very few contacts(the buried surface area is very similar for the two complexes,ca. 200 Å2); (4) the carboxyl group of NeuGc interacts with theguanidinium group of an arginine residue; and (5) the distinctivehydroxyl group of NeuGc is placed in a small hydrophilic pocket,where it may form three hydrogen bonds with the surroundingantibody residues. Despite these similarities, the orientation of theNeuGc-GM3 saccharide moiety in the binding site as well as theshape of the binding cavity are quite different in the two complexes.

3.5. Modelling the interaction between chP3 mAb and itsanti-idiotype 1E10

In order to investigate the antigen-mimicry effect exerted bythe P3 anti-idiotypic antibody 1E10, which has shown promisingresults as a therapeutic cancer vaccine, we decided to construct acomputer model of the 1E10 complex with the chP3 idiotype andcompare it with the modelled chP3–sugar complex. To this end, weused the crystal structure of the chP3 Fab with five modelled vari-ants of the tip of the VH CDR3 loop (Fig. 1C). For 1E10, no structuraldata were available, and we therefore resorted to construct, fromits sequence, a theoretical three-dimensional model of its variabledomain (Fv), using the WAM server (Whitelegg and Rees, 2000). The1E10 mAb has a relatively long VH CDR3, comprising 12 residues,which cannot be reliably modelled. To tackle this problem, whichcould affect the chP3–1E10 docking simulations, we generated 100different conformations of the VH CDR3, which after being clus-tered provided five different possible structures (Fig. 1D).

The docking simulations were performed using the programRosettaDock (version 2.0) (Gray et al., 2003). We combined the fiveFv structures generated by molecular modelling (in the case of chP3,based on the crystal structure) for each antibody in 25 independentsimulations. The results from each run were pooled and ranked bytheir energy scores, and the 20 top-scoring (lowest energy) decoyswere selected for further analysis.

The next step was to use the RosettaInterface (Kortemme andBaker, 2002) computational mutagenesis module to identify themodels that were consistent with the majority of the bindingresults available for the chP3 and 1E10 mutants. This program esti-mates ��G for the complex formation of the mutants comparedto the wild-type. All the VH single mutants tested experimen-tally (Arg31 → Ser, Arg98 → Ser, Arg100A → Ser, Arg98 → Thr andTrp100E → Tyr) were subjected to free energy calculations. Onlyfor one of the 20 top-scoring models, the calculated free energydifferences were in agreement with the experimental results. Thismodel is represented in Fig. 3(panels C and D). It shows VH Arg31 ofchP3, a residue previously shown to be important for 1E10 binding(López-Requena et al., 2007a), to be strongly involved in the inter-action with its anti-idiotype (Fig. 3D). This residue is located in anegatively charged pocket created by the side chains of residues

Asp33, Glu95 and the backbone oxygen of Asp96 from the 1E10heavy chain. The remaining chP3 VH residues probed by mutations,which did not affect the binding to 1E10, are found to be clearlypointing away from the binding site (Arg98 and Arg100A) or noteven close to it (Trp100E) (not shown).

3472 A. Talavera et al. / Molecular Immunology 46 (2009) 3466–3475

Fig. 3. Models of the chP3/NeuGc-GM3 and chP3/1E10 complexes. (A) Overview of the model of the chP3/NeuGc-GM3 trisaccharide complex. The chP3 surface is coloredby its electrostatic potential (+10 kT blue, 0 kT white and −10 kT red). The carbohydrate is shown in yellow sticks, with oxygen atoms in red and the nitrogen in blue. TheN-glycolyl group of NeuGc is buried in a small cavity. The distinctive hydroxyl oxygen of this group is marked with a black arrow head. (B) Detailed view of the NeuGc-GM3trisaccharide/chP3 interactions. The trisaccharide is colored as in A, while chP3 residues are colored in green (carbons), red (oxygens) and blue (nitrogens). Potential hydrogenbonds are indicated as black dashed lines. The N-glycolyl hydroxyl group is shown as a small red sphere. (C) Overall representation of the model of the chP3/1E10 complex.T E10 hei chaina of the

lo9(PdVT

e5fli

sTftea

he chP3 heavy and light chains are colored in green and wheat, respectively. The 1nteractions at the protein–protein interface. The color code is the same as in C. Sidend in black for chP3. Not all interacting residues are shown due to the complexity

Most of the contacts of 1E10 with its idiotypic ligand are estab-ished through its VH domain. The 1E10 VH CDR3 provides mostf the binding contacts, involving eight consecutive residues (from5 to 100B). VH CDR1 contributes with three amino acid residuesSer31, Tyr32, and Asp33), and four residues of the VH CDR2 (Trp50,he52, Asp54 and Lys58) are also involved in contacts. The VLomain contributes with only four amino acid residues, two fromL CDR3 (Asn92 and Thr93) and two from VL CDR1 (Ser30 andyr32).

The chP3 epitope recognized by 1E10 comprises 16 residues thatxclusively belong to the chP3 VH domain (1, 2, 26–28, 30–32, 53,4, 97–100A, 102). This observation is consistent with the resultsrom previous experiments, which demonstrated that the chP3ight chain could be replaced by the 1E10 light chain without affect-ng the binding to 1E10 (López-Requena et al., 2007b).

The calculated shape complementarity (sc) in our model isc = 0.57 and the average area buried upon interaction is 672 Å2.

he sc parameter is slightly lower than the values calculated forour other idiotype–anti-idiotype complexes deposited in the Pro-ein Data Bank (http://www.rcsb.org; Berman et al., 2000): 1iai (Bant al., 1994), 1pg7 (Eigenbrot et al., 2003), 1cic (Bentley et al., 1990)nd 1dvf (Braden et al., 1996). In these complexes, the area buried

avy chain is shown in blue, and its light chain in light blue. (D) Detailed view of thes at the chP3/1E10 interface are represented as sticks and labeled, in blue for 1E10figure.

upon interaction is ca. 1000 Å2, which is larger by approximatelyone-third compared to the modelled chP3/1E10 complex. This largedifference in interaction surface can be explained by the fact thatfor those four complexes, both the heavy and the light chains fromeach antibody contribute to the interaction, whereas our results aswell as previously reported experimental data (López-Requena etal., 2007a,b) suggest that chP3 and 1E10 mainly interact throughtheir heavy chains. A similar heavy chain/heavy chain interactionof an idiotype–anti-idiotype pair has recently been described forthe complex of the anti-HIV-1 antibody 2F5 with its anti-idiotype3H6 (3bqu; Bryson et al., 2008).

3.6. Estimation of binding affinities

Based on a non-competitive ELISA, we estimate the dissocia-tion constant (KD) of the chP3/1E10 binding to be in the order

of 10−8 M (data not shown). To assess the KD of the chP3/NeuGc-GM3 complex, we compared the recognition capacity of chP3 forL1210 cells, which express the NeuGc-GM3 ganglioside, with thatof the 14F7 antibody, which has a binding affinity of 2.5 × 10−8 M,as determined by ELISA (Rojas et al., 2004). In a competitive

A. Talavera et al. / Molecular Immun

Fig. 4. Competitive binding assay (chP3 versus ch14F7). The histograms, obtainedby FACS analysis, correspond to incubation of L1210 cells with different concentra-tions (0, 10, 100 and 1000 �g/ml) of chP3 and a constant concentration (10 �g/ml)of biotinylated ch14F7. Cell staining was measured using fluorescein isothiocyanate(FITC)-conjugated streptavidin. The histogram at the bottom, filled in gray, corre-sponds to incubation of L1210 cells with a non-binding biotinylated antibody andFITC-conjugated streptavidin (negative control). The X-axis corresponds to fluores-cfm

flttctt1

ptspi

3

hpctatm

ito1wTmcAat

(also known as the VHQ52 family) are commonly used by anti-ganglioside antibodies (Boffey et al., 2005), suggesting that thespecificities of many ganglioside-binding immunoglobulins aredefined by a limited group of VH genes, encoding particular amino

Fig. 5. Is there a mimicry of the NeuGc-GM3 ganglioside by the 1E10 antibody? Thepredicted models of the chP3/NeuGc-GM3 and chP3/1E10 complexes are superim-posed. The chP3 VH CDR loops (from the chP3/NeuGc-GM3 complex) are shown

ence intensity. The dotted line represents the experimentally determined cut-offor cells recognized by 14F7. The percentage of antibody-stained cells as well as the

ean fluorescence intensity (MFI) are given for each histogram.

uorescence-activated cell sorting (FACS) assay, chP3 was not ableo inhibit the binding of the chimeric 14F7 antibody (ch14F7)o L1210 cells, even when using a 100-fold higher concentrationompared to ch14F7 (Fig. 4). Based on these results, we estimatehe binding affinity of the chP3/NeuGc-GM3 couple to be at leastwo orders of magnitude lower than that of the chP3/1E10 and4F7/NeuGc-GM3 complexes.

The lower binding affinity of chP3 towards NeuGc-GM3 com-ared to ch14F7 is probably a result of the fewer hydrogen bondso the carbohydrate moiety of the ganglioside, combined with thehape of the binding site. According to our models, the bindingocket that accommodates the NeuGc sialic acid is more shallow

n chP3 compared to the binding cleft of 14F7.

.7. NeuGc-GM3 mimicry by 1E10

The concept of antigen mimicry by anti-idiotypic antibodiesas long been discussed (Strosberg et al., 1985). One of the pur-oses behind the construction of the computer models of thehP3/NeuGc-GM3 and chP3/1E10 complexes, was to determine ifhere are structural similarities between the carbohydrate antigennd the 1E10 anti-idiotype, which are important for the recogni-ion of these molecules by the P3 antibody. Comparison of the two

odels, however, suggests that this is not the case.The major part of the sialic acid residue in NeuGc-GM3, namely

ts N-glycolyl group, its 4-hydroxyl group and the glycerol tail, bindo a pocket that remains empty upon 1E10 binding, according tour models (Fig. 5). Blocking of NeuGc-GM3 binding to chP3 byE10 is due mainly to the strong overlap of the 1E10 VL domainith both the galactose and glucose residues of the ganglioside.

he geometry of the tip of the chP3 VH CDR3 differs in the two

odels, although the loop conformation in the chP3/NeuGc-GM3

omplex is also compatible with 1E10 binding. Furthermore, VHrg31 in chP3, which plays a key role for ganglioside binding, adoptsdifferent conformation in the complex with 1E10, although in

he latter complex this arginine residue also interacts with nega-

ology 46 (2009) 3466–3475 3473

tively charged oxygens, of an aspartic and a glutamic acid residue,respectively (see Fig. 3B and D).

Nevertheless, according to our models, most of the P3 residuesthat interact with the NeuGc-GM3 antigen also participate in inter-actions with the anti-idiotypic antibody 1E10. These residues (allin the P3 VH domain) include: Arg31 and Tyr32 in VH CDR1, Gly53and Gly54 in VH CDR2 (not labeled in Fig. 3), and the 5-residuesegment Val97-Arg100A at the tip of VH CDR3.

The ability of 1E10 to elicit antibodies (so-called ‘Ab3’) that rec-ognize the NeuGc-GM3 ganglioside can hence not be explained interms of structural mimicry. We hypothesize that some of the Ab3antibodies generated upon immunization with 1E10 should bindthis antibody in a very similar manner as P3. Therefore, antibod-ies belonging to this “P3-like” Ab3 subset should have a bindingsite topography that resembles the P3 binding site. Some of these“P3-like” Ab3 mAbs would then also be capable of recognizing theNeuGc-GM3 ganglioside. This similarity in binding site topographybetween P3 and some of the Ab3 mAbs would most likely be accom-panied with a high amino acid sequence similarity, at least betweenthe VH CDRs.

We previously reported that P3 is an antibody of germlineorigin (Perez et al., 2001). An immunogenetic analysis of theP3 VH nucleotide sequence performed in this work, using thetools implemented with the IMGT/LIGM-DB database (Giudicelliet al., 2006) and IMGT/V-QUEST (Brochet et al., 2008) atIMGT, the international ImMunoGeneTics information system(http://www.imgt.org; Lefranc et al., 2009), revealed 100% identityof the P3 V gene-coded region with the IGHV2-6-4*01 germlinegene. It has been reported that genes of this IGHV2 subgroup

in green. The trisaccharide is shown in yellow, with oxygen atoms in red and thenitrogen in blue. The 1E10 binding determinants close to the trisaccharide bind-ing site are represented as sticks, colored in light and dark blue for the light andheavy chains, respectively. Note that there are no 1E10 amino acid residues in thecavity occupied by the N-glycolyl group of NeuGc, which argues against structuralmimicry.

3 Immu

aonsswev

4

ce

A

wvF

R

A

A

B

B

B

B

B

B

B

B

C

C

C

C

C

D

E

474 A. Talavera et al. / Molecular

cid sequences of VH CDR1 and CDR2. This fact further supportsur reasoning that the P3-like Ab3 mAbs resulting from immu-ization with the 1E10 anti-idiotype should have a high sequenceimilarity with P3, since they probably are encoded by genes of aimilar human VH subgroup. Testing the validity of this hypothesisill require sequencing the variable regions of the Ab3 antibodies

licited in cancer patients that have been treated with the 1E10accine.

. Accession numbers

The atomic coordinates and experimental structure factors ofhP3 Fab have been deposited with the Protein Data Bank (Bermant al., 2000) under accession number 3IU4.

cknowledgements

We thank the staff at the ESRF, Grenoble, for support. This workas funded by the Center for Molecular Immunology, the Uni-

ersity of Oslo as well as by grants from the Norwegian Canceroundation.

eferences

dams, P.D., Grosse-Kunstleve, R.W., Hung, L.W., Ioerger, T.R., McCoy, A.J., Mori-arty, N.W., Read, R.J., Sacchettini, J.C., Sauter, N.K., Terwilliger, T.C., 2002. Phenix:building new software for automated crystallographic structure determination.Acta Cryst. D58, 1948–1954.

lfonso, M., Díaz, A., Hernández, A.M., Pérez, A., Rodríguez, E., Bitton, R., Pérez,R., Vázquez, A.M., 2002. An anti-idiotype vaccine elicits a specific responseto N-glycolyl sialic acid residues of glycoconjugates in melanoma patients. J.Immunol. 168, 2523–2529.

an, N., Escobar, C., Garcia, R., Hasel, K., Day, J., Greenwood, A., McPherson, A., 1994.Crystal structure of an idiotype–anti-idiotype Fab complex. Proc. Natl. Acad. Sci.U.S.A. 91, 1604–1608.

eatty, J.D., Beatty, B.G., Vlahos, W.G., 1987. Measurement of monoclonal antibodyaffinity by non-competitive enzyme immunoassay. J. Immunol. Methods 100,173–179.

entley, G.A., Boulot, G., Riottot, M.M., Poljak, R.J., 1990. Three-dimensional structureof an idiotope–anti-idiotope complex. Nature 348, 254–257.

erman, H.M., Westbrook, J., Feng, Z., Gilliland, G., Bhat, T.N., Weissig, H., Shindyalov,I.N., Bourne, P.E., 2000. The Protein Data Bank. Nucleic Acids Res. 28, 235–242.

offey, J., Odaka, M., Nicoll, D., Wagner, E.R., Townson, K., Bowes, T., Conner,J., Furukawa, K., Willison, H.J., 2005. Characterisation of the immunoglobu-lin variable region gene usage encoding the murine anti-ganglioside antibodyrepertoire. J. Neuroimmunol. 165, 92–103.

raden, B.C., Fields, B.A., Ysern, X., Dall’Acqua, W., Goldbaum, F.A., Poljak, R.J., Mar-iuzza, R.A., 1996. Crystal structure of an Fv–Fv idiotope–anti-idiotope complexat 1.9 Å resolution. J. Mol. Biol. 264, 137–151.

rochet, X., Lefranc, M.P., Giudicelli, V., 2008. IMGT/V-QUEST: the highly customizedand integrated system for IG and TR standardized V-J and V-D-J sequence anal-ysis. Nucleic Acids Res. 36, W503–W508.

ryson, S., Julien, J.-P., Isenman, D.E., Kunert, R., Katinger, H., Pai, E.F., 2008. Crystalstructure of the complex between the F′

ab fragment of the cross-neutralizinganti-HIV-1 antibody 2F5 and the Fab fragment of its anti-idiotypic antibody 3H6.J. Mol. Biol. 382, 910–919.

arr, A., Mullet, A., Mazorra, Z., Vázquez, A.M., Alfonso, M., Mesa, C., Rengifo, E., Pérez,R., Fernández, L.E., 2000. A mouse IgG1 monoclonal antibody specific for N-glycolyl GM3 ganglioside recognized breast and melanoma tumors. Hybridoma19, 241–247.

ollaborative Computational Project, Number 4, 1994. The CCP4 suite: programs forprotein crystallography. Acta Cryst. D50, 760–763.

hen, L., Chang, S., Mohan, C., 2002. Molecular signatures of antinuclearantibodies—contributions of heavy chain CDR residues. Mol. Immunol. 39,333–347.

occa, B.A., Seal, S.N., D’Agnillo, P., Mueller, Y.M., Katsikis, P.D., Rauch, J.,Weigert, M., Radic, M.Z., 2001. Structural basis for autoantibody recognition ofphosphatidylserine-beta2 glycoprotein I and apoptotic cells. Proc. Natl. Acad.Sci. U.S.A. 98, 13826–13831.

oloma, M.J., Hastings, A., Wims, L.A., Morrison, S.L., 1992. Novel vectors for theexpression of antibody molecules using variable regions generated by poly-merase chain reaction. J. Immunol. Methods 152, 89–104.

íaz, A., Alfonso, M., Alonso, R., Saurez, G., Troche, M., Catalá, M., Díaz, R.M., Pérez, R.,Vázquez, A.M., 2003. Immune responses in breast cancer patients immunizedwith an anti-idiotype antibody mimicking NeuGc-containing gangliosides. Clin.Immunol. 107, 80–89.

igenbrot, C., Meng, Y.G., Krishnamurthy, R., Lipari, M.T., Presta, L., Devaux, B., Wong,T., Moran, P., Bullens, S., Kirchhofer, D., 2003. Structural insight into how an

nology 46 (2009) 3466–3475

anti-idiotypic antibody against D3H44 (anti-tissue factor antibody) restoresnormal coagulation. J. Mol. Biol. 331, 433–446.

Emsley, P., Cowtan, K., 2004. Coot: model-building tools for molecular graphics. ActaCryst. D60, 2126–2132.

Engh, R.A., Huber, R., 1991. Accurate bond and angle parameters for X-ray proteinstructure refinement. Acta Cryst. A47, 392–400.

Evans, P., 2006. Scaling and assessment of data quality. Acta Cryst. D62, 72–82.Giles, I.P., Haley, J.D., Nagl, S., Isenberg, D.A., Latchman, D.S., Rahman, A., 2003.

A systematic analysis of sequences of human antiphospholipid and anti-�2-glycoprotein I antibodies: the importance of somatic mutations and certainsequence motifs. Semin. Arthritis Rheum. 32, 246–265.

Giles, I., Lambrianides, N., Latchman, D., Chen, P., Chukwuocha, R., Isenberg, D., Rah-man, A., 2005. The critical role of arginine residues in the binding of humanmonoclonal antibodies to cardiolipin. Arthritis Res. Ther. 7, R47–R56.

Giudicelli, V., Duroux, P., Ginestoux, C., Folch, G., Jabado-Michaloud, J., Chaume,D., Lefranc, M.-P., 2006. IMGT/LIGM-DB, the IMGT® comprehensive databaseof immunoglobulin and T cell receptor nucleotide sequences. Nucleic Acids Res.34, D781–D784.

Gray, J.J., Moughon, S., Wang, C., Schueler-Furman, O., Kuhlman, B., Rohl, C.A., Baker,D., 2003. Protein–protein docking with simultaneous optimization of rigid-bodydisplacement and side-chain conformations. J. Mol. Biol. 331, 281–299.

Guth, A.M., Zhang, X., Smith, D., Detanico, T., Wysocki, L.J., 2003. Chromatin speci-ficity of anti-double-stranded DNA antibodies and a role for Arg residues in thethird complementarity-determining region of the heavy chain. J. Immunol. 171,6260–6266.

Guthmann, M.D., Castro, M.A., Cinat, G., Venier, C., Koliren, L., Bitton, R.J., Vázquez,A.M., Fainboim, L., 2006. Cellular and humoral immune response to N-glycolyl-GM3 elicited by prolonged immunotherapy with an anti-idiotypic vaccinein high-risk and metastatic breast cancer patients. J. Immunother. 29, 215–223.

Hernández, A.M., Toledo, D., Martínez, D., Grinán, T., Brito, V., Macías, A., Alfonso,S., Rondón, T., Suárez, E., Vázquez, A.M., Pérez, R., 2008. Characterization of theantibody response against NeuGcGM3 ganglioside elicited in non-small cell lungcancer patients immunized with an anti-idiotype antibody. J. Immunol. 181,6625–6634.

Humphrey, W., Dalke, A., Schulten, K., 1996. VMD: visual molecular dynamics. J. Mol.Graph. 14, 33–38.

James, L.C., Hale, G., Waldman, H., Bloomer, A.C., 1999. 1.9 Å structure of the thera-peutic antibody CAMPATH-1H Fab in complex with a synthetic peptide antigen.J. Mol. Biol. 289, 293–301.

Kammann, M., Laufs, J., Schell, J., Gronenborn, B., 1989. Rapid insertional mutagen-esis of DNA by polymerase chain reaction (PCR). Nucleic Acids Res. 17, 5404.

Kannagi, R., Yin, J., Miyazaki, K., Izawa, M., 2008. Current relevance of incompletesynthesis and neo-synthesis for cancer-associated alteration of carbohydratedeterminants—Hakomori’s concepts revisited. Biochim. Biophys. Acta 1780,525–531.

Kantardjieff, K.A., Rupp, B., 2003. Matthews coefficient probabilities: improved esti-mates for unit cell contents of proteins, DNA, and protein–nucleic acid complexcrystals. Protein Sci. 12, 1865–1871.

Kortemme, T., Baker, D., 2002. A simple physical model for binding energy hot spotsin protein–protein complexes. Proc. Natl. Acad. Sci. U.S.A. 99, 14116–14121.

Kortemme, T., Joachimiak, L.A., Bullock, A.N., Schuler, A.D., Stoddard, B.L., Baker, D.,2004. Computational redesign of protein–protein interaction specificity. Nat.Struct. Mol. Biol. 11, 371–379.

Krengel, U., Olsson, L.-L., Martínez, C., Talavera, A., Rojas, G., Mier, E., Ångström, J.,Moreno, E., 2004. Structure and molecular interactions of a unique antitumorantibody specific for N-glycolyl GM3. J. Biol. Chem. 279, 5597–5603.

Laskowski, R.A., MacArthur, M.W., Moss, D.S., Thornton, J.M., 1993. PROCHECK: aprogram to check the stereochemical quality of protein structures. J. Appl. Crys-tallogr. 26, 283–291.

Lefranc, M.P., Giudicelli, V., Ginestoux, C., Jabado-Michaloud, J., Folch, G., Bellahcene,F., Wu, Y., Gemrot, E., Brochet, X., Lane, J., Regnier, L., Ehrenmann, F., Lefranc, G.,Duroux, P., 2009. IMGT, the international ImMunoGeneTics information system.Nucleic Acids Res. 37, D1006–D1012.

Lesinski, G.B., Westerink, M.A.J., 2001. Novel vaccine strategies to T-independentantigens. J. Microbiol. Methods 47, 135–149.

Leslie, A.G.W., 1992. Recent changes to the MOSFLM package for processing film andimage plate data. Joint CCP4 + ESF-EACBM Newsletter on Protein Crystallogra-phy No. 26.

Leslie, A.G.W., 2006. The integration of macromolecular diffraction data. Acta Cryst.D62, 48–57.

López-Requena, A., Mateo de Acosta, C., Pérez, A., Valle, A., Lombardero, J., Sosa,K., Pérez, R., Vázquez, A.M., 2003. Chimeric anti-N-glycolyl-ganglioside andits anti-idiotypic mAbs: immunodominance of their variable regions. Hybrid.Hybridomics 22, 235–243.

López-Requena, A., Mateo de Acosta, C., Moreno, E., González, M., Puchades, Y.,Talavera, A., Vispo, N.S., Vázquez, A.M., Pérez, R., 2007a. Gangliosides, Ab1 andAb2 antibodies. I. Towards a molecular dissection of an idiotype–anti-idiotypesystem. Mol. Immunol. 44, 423–433.

López-Requena, A., Rodríguez, M., Mateo de Acosta, C., Moreno, E., Puchades, Y.,

González, M., Talavera, A., Valle, A., Hernández, T., Vázquez, A.M., Pérez, R.,2007b. Gangliosides, Ab1 and Ab2 antibodies. II. Light versus heavy chain: anidiotype–anti-idiotype case study. Mol. Immunol. 44, 1015–1028.

Malykh, Y.N., Krisch, B., Shaw, L., Warner, T.G., Sinicropi, D., Smith, R., Chang,J., Schauer, R., 2001. Distribution and localization of CMP-N-acetylneuraminicacid hydroxylase and N-glycolylneuraminic acid-containing glycoconjugates in

Immun

M

MM

M

M

M

N

P

P

R

R

R

S

S

S

A. Talavera et al. / Molecular

porcine lymph node and peripheral blood lymphocytes. Eur. J. Cell Biol. 80,48–58.

arquina, G., Waki, H., Fernandez, L.E., Kon, K., Carr, A., Valiente, O., Perez, R.,Ando, S., 1996. Gangliosides expressed in human breast cancer. Cancer Res. 56,5165–5171.

atthews, B.W., 1968. Solvent content of protein crystals. J. Mol. Biol. 33, 491–497.cCoy, A.J., 2007. Solving structures of protein complexes by molecular replacement

with Phaser. Acta Cryst. D63, 32–41.oreno, E., Lanne, B., Vázquez, A.M., Kawashima, I., Tai, T., Fernández, L.E., Karls-

son, K.-A., Ångström, J., Pérez, R., 1998. Delineation of the epitope recognizedby an antibody specific for N-glycolylneuraminic acid-containing gangliosides.Glycobiology 8, 695–705.

orris, G.M., Goodsell, D.S., Halliday, R.S., Huey, R., Hart, W.E., Belew, R.K., Olson,A.J., 1998. Automated docking using a Lamarckian genetic algorithm and anempirical binding free energy function. J. Comput. Chem. 19, 1639–1662.

urshudov, G.N., Vagin, A.A., Dodson, E.J., 1997. Refinement of macromolecularstructures by the maximum-likelihood method. Acta Cryst. D53, 240–255.

eninger, E., Díaz, R.M., de la Torre, A., Rives, R., Díaz, A., Saurez, G., Gabri, M.R.,Alonso, D.F., Wilkinson, B., Alfonso, A.M., Combet, T., Pérez, R., Vázquez, A.M.,2007. Active immunotherapy with 1E10 anti-idiotype vaccine in patients withsmall cell lung cancer: report of a phase I trial. Cancer Biol. Ther. 6, 145–150.

erez, A., Lombardero, J., Mateo, C., Mustelier, G., Alfonso, M., Vazquez, A.M., Perez,R., 2001. Immunogenetic analysis of variable regions encoding AB1 and �-typeAB2 antibodies from the NeuGc-containing ganglioside family. Hybridoma 20,211–221.

errakis, A., Morris, R., Lamzin, V.S., 1999. Automated protein model building com-bined with iterative structure refinement. Nat. Struct. Biol. 6, 458–463.

odríguez, M., Llanes, L., Pérez, A., Pérez, R., Vázquez, A.M., 2003. Genera-tion and characterization of an anti-idiotype monoclonal antibody related toGM3(NeuGc) ganglioside. Hybrid. Hybridomics 22, 307–314.

odríguez, M., Roque-Navarro, L., López-Requena, A., Moreno, E., Mateo de Acosta,C., Pérez, R., Vázquez, A.M., 2007. Insights into the immunogenetic basis of twoganglioside-associated idiotypic networks. Immunobiology 212, 57–70.

ojas, G., Talavera, A., Munoz, Y., Rengifo, E., Krengel, U., Ångström, J., Gavilondo,J., Moreno, E., 2004. Light-chain shuffling results in successful phage displayselection of functional prokaryotic-expressed antibody fragments to N-glycolylGM3 ganglioside. J. Immunol. Methods 293, 71–83.

ˇali, A., Blundell, T.L., 1993. Comparative protein modelling by satisfaction of spatialrestraints. J. Mol. Biol. 234, 779–815.

anger, F., Nicklen, S., Coulson, A.R., 1977. DNA sequencing with chain-terminatinginhibitors. Proc. Natl. Acad. Sci. U.S.A. 74, 5463–5467.

iebert, H.-C., Reuter, G., Schauer, R., von der Lieth, C.-W., Dabrowski, J., 1992. Solu-tion conformations of GM3 gangliosides containing different sialic acid residuesas revealed by NOE-based distance mapping, molecular mechanics, and molec-ular dynamics calculations. Biochemistry 31, 6962–6971.

ology 46 (2009) 3466–3475 3475

Strosberg, A.D., Guillet, J.G., Chamat, S., Hoebeke, J., 1985. Recognition of physiolog-ical receptors by anti-idiotypic antibodies: molecular mimicry of the ligand orcross-reactivity? Curr. Top. Microbiol. Immunol. 119, 91–110.

Stults, C.L.M., Sweeley, C.C., Macher, B.A., 1989. Glycosphingolipids: structure, bio-logical source, and properties. Methods Enzymol. 179, 167–214.

Teodoro, M.L., Kavraki, L.E., 2003. Conformational flexibility models for the receptorin structure based drug design. Curr. Pharm. Des. 9, 1635–1648.

Vázquez, A.M., Alfonso, M., Lanne, B., Karlsson, K.-A., Carr, A., Barroso, O., Fernández,L.E., Rengifo, E., Lanio, M.E., Alvarez, C., Zeuthen, J., Pérez, R., 1995. Generation of amurine monoclonal antibody specific for N-glycolylneuraminic acid-containinggangliosides that also recognizes sulfated glycolipids. Hybridoma 14, 551–556.

Vázquez, A.M., Pérez, A., Hernández, A.M., Macías, A., Alfonso, M., Bombino, G., Pérez,R., 1998. Syngeneic anti-idiotypic monoclonal antibodies to an anti-NeuGc-containing ganglioside monoclonal antibody. Hybridoma 17, 527–534.

Whitelegg, N.R.J., Rees, A.R., 2000. WAM: an improved algorithm for modelling anti-bodies on the WEB. Protein Eng. 13, 819–824.

Zwart, P.H., Afonine, P.V., Grosse-Kunstleve, R.W., Hung, L.-W., Ioerger, T.R., McCoy,A.J., McKee, E., Moriarty, N.W., Read, R.J., Sacchettini, J.C., Sauter, N.K., Storoni,L.C., Terwilliger, T.C., Adams, P.D., 2008. Automated structure solution with thePHENIX suite. Methods Mol. Biol. 426, 419–435.

Glossary

CDR: complementarity determining regionELISA: enzyme-linked immunosorbent assayFab: antigen-binding fragments of immunoglobulinsFACS: fluorescence-activated cell sortingFc: fragment, crystallizable (Fab constant domain)Fv: Fab variable domainGal: galactoseGlc: glucoseI: intensityIg: immunoglobulinmAb: monoclonal antibodyNeuAc: 5′-N-acetylneuraminidateNeuGc: 5′-N-glycolylneuraminidateRMSD: root mean square deviation

sc: shape complementarityVH: heavy chain variable domainVH: heavy chain V gene segmentVL: light chain variable domain