Aw

SP

a

ARRAA

KICCPCAC

1

tsamrwflbtoemdftest

0d

Molecular Immunology 51 (2012) 66–72

Contents lists available at SciVerse ScienceDirect

Molecular Immunology

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

tomic resolution model of the antibody Fc interactionith the complement C1q component

ebastian Schneider, Martin Zacharias ∗

hysik-Department T38, Technische Universität München, James Franck Str. 1, 85748 Garching, Germany

r t i c l e i n f o

rticle history:eceived 5 December 2011eceived in revised form 17 January 2012ccepted 6 February 2012vailable online 15 March 2012

eywords:gG–C1q interactionomplement activation

a b s t r a c t

The globular C1q heterotrimer is a subunit of the C1 complement factor. Binding of the C1q subunit to theconstant (Fc) part of antibody molecules is a first step and key event of complement activation. Althoughthree-dimensional structures of C1q and antibody Fc subunits have been determined experimentally noatomic resolution structure of the C1q–Fc complex is known so far. Based on systematic protein–proteindocking searches and Molecular Dynamics simulations a structural model of the C1q–IgG1–Fc-bindinggeometry has been obtained. The structural model is compatible with available experimental data onthe interaction between the two partner proteins. It predicts a binding geometry that involves mainlythe B-subunit of the C1q-trimer and both subunits of the IgG1–Fc-dimer with small conformational

1q–Fc structure modelingrotein–protein dockingomplement component C1ntibody design1 globular domain

adjustments with respect to the unbound partners to achieve high surface complementarity. In additionto several charge–charge and polar contacts in the rim region of the interface it also involves nonpolarcontacts between the two proteins and is compatible with the carbohydrate moiety of the Fc subunit. Themodel for the complex structure provides a working model for rationalizing available biochemical dataon this important interaction and can form the basis for the design of Fc variants with a greater capacityto activate the complement system for example on binding to cancer cells or other target structures.

. Introduction

The C1 component of the complement system is a heteropen-americ complex of five multi-domain proteins, a recognitionubunit C1q and two C1r and two C1s serine proteases. It triggersctivation of the classical pathway of complement which is aajor part of the innate immunity (Ehrnthaller et al., 2011). The

ecognition component of the C1 complex is the C1q moleculehich consists of six subunits with an overall shape of a bouquet ofowers (illustrated in Fig. 1). Complement activation is triggeredy the initial binding of the globular domains of the C1q subunitso the Fc portion of IgG or IgM antibodies bound to antigensn the bacterial surface (Sarma and Ward, 2011; Ehrnthallert al., 2011). Multiple binding is required to stabilize this initialolecular complex and leads to a cooperative response which can

istinguish between antibody molecules bound to a bacteriumrom those free in solution. The process of multiple C1q bindingo a bacterium eventually results in complement activation and

limination of the bacterium. Stable binding of two or more of theix globular heads of C1q has to be established in order to activatehe complement C1 component. The binding event triggers the

∗ Corresponding author. Tel.: +49 89 289 12335; fax: +49 89 289 12444.E-mail address: martin.zacharias@ph.tum.de (M. Zacharias).

161-5890/$ – see front matter © 2012 Elsevier Ltd. All rights reserved.oi:10.1016/j.molimm.2012.02.111

© 2012 Elsevier Ltd. All rights reserved.

activation of the C1r and C1s protease subunits of C1 which, inturn, activates the classical complement cascade (Zlatarova et al.,2006; Trouw and Daha, 2011; Ehrnthaller et al., 2011).

The C1q globular domain is a heterotrimeric complex withan arrangement of protein modules that has been found in sev-eral other proteins (Ghai et al., 2007). The C1q globular domaindoes not only bind to antibodies but also to a variety of otherligands including C-reactive protein, pentraxin 3 and several bac-terial surface proteins (Lu et al., 2008). The structure of the C1qsubunit (Gaboriaud et al., 2003, 2004) and the structure of anti-body Fc regions (Saphire et al., 2001; Radaev et al., 2001) havebeen determined experimentally. In addition, residues importantfor the interaction of the two proteins were identified by muta-genesis studies (e.g. Kishore et al., 2004; Kojouharova et al., 2003,2004; Moore et al., 2010) and an approximate arrangement wassuggested (Duncan and Winter, 1988; Gaboriaud et al., 2003). How-ever, the complex structure between the two proteins has so farnot been determined experimentally in atomic detail. In recentyears, putative interaction regions between C1q and the antibodyFc domain have been investigated by mutagenesis studies with theaim to enhance the complement activation of therapeutic anti-

bodies (Idusogie et al., 2000; Moore et al., 2010). Engineeringthe antibody Fc region to enhance the cytotoxic activity of anti-bodies offers the potential of designing monoclonal therapeuticantibodies with greatly increased destructive capabilities against

S. Schneider, M. Zacharias / Molecular

Fig. 1. Schematic view of the complement factor C1q (without the enzymatic com-pFi

spdaiWoeitbIftMsagb

2

2

uaa1mwpu

onents: C1r2C1s2) and the interaction of the C1q globular head domain with thec domain (in red) of an IgG antibody. (For interpretation of the references to colorn this figure legend, the reader is referred to the web version of this article.)

pecific target cells. Such target cells include not only invadingathogens but also cancer cells. The availability of mutagenesisata on residues that may participate in the interaction betweenntibody Fc and the C1q head domains offer also the possibil-ty to generate and evaluate structural models for the interaction.

e have employed computational docking to generate a modelf the interaction geometry for the complex including availablexperimental data on the binding region. The structural models compatible with all available mutagenesis data and explainshe experimental observation that antibodies with only one intactinding motif are still able to bind C1q (Michaelson et al., 2006).

n addition, the model assigns for each of the residues, that wereound to be important for binding, a defined interaction partner onhe other protein. The docked complex was refined by employing

olecular Dynamics (MD) simulations in explicit solvent and canerve as a working model to better understand the details of thentibody IgG1–Fc interaction with C1q and to plan future muta-enesis studies for refining the structural model or for improvinginding specificity and affinity.

. Materials and methods

.1. Protein–protein docking using ATTRACT

The protein data bank (pdb) entries 1PK6 and 1HZH served asnbound C1q and antibody Fc partner structures, respectively, forll protein–protein docking searches. Only the IgG1 Fc part of thentibody structure (residues 245–475 of chains H and K) in theHZH-entry was used for docking. The compatibility of the docked

odel structures with the placement of the mobile Fab fragmentsas considered separately (see Section 3). The ATTRACT dockingrogram (Zacharias, 2003, 2010; Fiorucci and Zacharias, 2010) wassed for systematic docking searches. It employs a coarse-grained

Immunology 51 (2012) 66–72 67

representation for protein binding partners with two pseudo atomsper residue representing the main chain (located at the back-bone nitrogen and backbone oxygen atoms, respectively). Smallamino acid side chains (Ala, Asp, Asn, Cys, Ile, Leu, Pro, Ser, Thr,Val) are represented by one pseudo atom (geometric mean ofside chain heavy atoms). Larger and more flexible side chains arerepresented by two pseudo atoms to account for the shape anddual chemical character of some side chains. Effective interactionsbetween pseudo-atoms are described by soft distance-dependentLennard–Jones (LJ)-type potentials (for details see Fiorucci andZacharias, 2010). A Coulomb type term accounts for electrostaticinteractions between real charges (Lys, Arg, Glu, Asp) damped bya distance-dependent dielectric constant (ε = 15r). The ATTRACTprotein–protein docking program was among the top perform-ing docking programs in the community wide docking challengeCAPRI (Critical Assessment of Predicted Interactions, Lensink andWodak, 2010). Furthermore, in a recent study we could show thatthe ATTRACT approach is particularly strong in case of dockingunbound protein structures if there is experimental data on thebinding data available (Schneider and Zacharias, 2011) indicatingthe reliability of the approach.

It is also possible to add interaction weights on residues thatare known to participate in binding, in order to bias the searchtowards regions that include residues important for interaction.For the present docking searches a weight of 1.5 for the attrac-tive interactions was included for residues that have been shownexperimentally to be important for C1q–Fc interaction. This typeof interaction weighting has been found previously on many testcases as the optimal bias to include experimental data on proteinbinding sites (Schneider and Zacharias, 2011). Systematic dock-ing was started from several starting points (spaced by ∼8 A) andorientations of the C1q partner near the roughly known bindingregion of the antibody Fc subunit. Each docking run consisted ofa set of energy minimizations in translational and orientationalvariables following published protocols (Zacharias, 2003; May andZacharias, 2008). Clustering of solutions was performed startingfrom the lowest energy complexes (best scoring) using an Rmsdlig(root mean square deviation of the ligand protein after superpo-sition of receptor proteins) cutoff between any two solutions of5 A. Comparison of known structures of the Fc portion of IgG anti-bodies in complex with different partner proteins indicates thatglobal adjustments may play a role during interaction. To accountfor such global conformational changes in the Fc segment duringdocking we employed the option of including energy minimizationin soft global collective modes in the docking ATTRACT program(May and Zacharias, 2005, 2008). The soft global modes for the Fc-segment were calculated using an elastic network model based onthe approach by Hinsen (1998).

2.2. Refinement of docked complexes at atomic resolution

Atomic models of the docked complexes were obtained bysuperposition of the atomic resolution partner structures onto thedocked complexes obtained at coarse-grained resolution. All fur-ther refinement steps were performed with the Amber11 all-atommolecular modeling package employing the parmff03.r1 force field(Case et al., 2010). In order to remove possible sterical overlapthe structures were first energy minimized (1500 steps, using adistance-dependent dielectric constant: ε(r) = 4r, r is the distancebetween atoms). The structures of complexes were further opti-mized using three consecutive MD simulations at temperatures of550 K (6 ps), 400 K (4 ps) and 300 K (4 ps) followed by 1500 EM steps.

During the MD-simulations harmonic distance restraints betweenC� atoms within each partner protein were applied. This allows fullflexibility of side chains and limited flexibility of the backbone ofeach partner and flexible adjustment of the interacting partners.

6 ecular

Sitetm

3

3

wseZttAFsIefinmorOaafsiFtofEtaZweomdtoKtibiIfsZfsdATs

8 S. Schneider, M. Zacharias / Mol

ubsequently, the complexes were solvated in octahedral boxesncluding explicit water molecules. After heating up during 1 nso 300 K and removal of positional restraints on the two proteins,ach docked complex was equilibrated for 11 ns. The final struc-ures were again energy minimized (2000 EM steps) and served as

odel structures for the C1q–Fc-complex.

. Results and discussion

.1. Systematic C1q–antibody-Fc docking

The putative interaction region of the globular C1q head trimerith the IgG1–Fc domain of antibodies has been characterized by

everal mutagenesis studies (Kaul and Loos, 1997; Kojouharovat al., 2003, 2004; Kishore et al., 2002, 2004; Roumenina et al., 2006;latarova et al., 2006; Gadjeva et al., 2008). It was possible to iden-ify key residues in C1q that when substituted significantly reducehe binding affinity to Fc (Fig. 2). These residues include ArgA162,rgB108, HisB117, ArgB129, ArgB163 and ArgC156. Several mutations inc were identified that influence binding to C1q including also someubstitutions that increase the affinity (Thommesen et al., 2000;dusogie et al., 2001; Presta, 2008; Michaelson et al., 2006; Mooret al., 2010). The Fc part consists of two identical chains. In theollowing the chain identity (A,B) will only be distinguished if wendicate a specific residue of one chain in the model structure (e.g.ot if an experimentally described mutation is discussed since theutation is present in both chains). For example, the substitution

f Asp270, Lys322 and especially Pro329 and Pro331 on Fc is known toeduce C1q binding (Burton et al., 1980; Idusogie et al., 2000, 2001;ganesyan et al., 2008). It has also been possible to increase theffinity of C1q to the Fc region of IgG1 after substitution of Lys326Trpnd Glu333Ser (Idusogie et al., 2001). In addition, Moore et al. (2010)ound that the substitution of Ser267Glu, His268Phe and Ser324Thrignificantly increased the binding affinity of C1q to Fc. Interest-ngly, the putative interaction regions on the surface of C1q andc show also a high degree of electrostatic complementarity (illus-rated in Fig. 2). The knowledge of the putative C1q binding regionn the Fc dimer allowed focussing the docking search by startingrom six initial placements close to the putative binding region.ach starting placement included ∼300 different starting orienta-ions that were docked using the ATTRACT energy minimizationpproach (Zacharias, 2003; May and Zacharias, 2008; Fiorucci andacharias, 2010). During docking the flexibility of the Fc proteinas approximately included by energy minimization in the 5 soft-

st normal modes obtained from an elastic network model (ENM)f each partner protein (see Section 2). The simultaneous opti-ization in translation and orientation and in the softest collective

egrees of freedom allowed for an induced fit during docking. Puta-ive residues involved in binding to Fc have also been investigatedn C1q by mutagenesis studies (Kojouharova et al., 2003, 2004;ishore et al., 2004; Roumenina et al., 2006) which resulted in

he identification of several basic (mainly Arg and His) residuesn the B as well as C subunits of the C1q heterotrimer important forinding (Fig. 2). The ATTRACT program allows one to include exper-

mental data on putative interaction regions as force field weights.nclusion of such data can significantly enhance the docking per-ormance resulting in improved ranking of near-native dockingolutions and more realistic docking geometries (Schneider andacharias, 2011). During docking force field weights for attractiveorces were doubled for several residues involved in binding asuggested by experimental mutagenesis studies. For the IgG1–Fc

omain this included the following residues: Pro329, Pro331, Lys322,sp270 and for the C1q molecule residues ArgB114, HisB117, ArgB129.he importance of these residues for binding was shown in severaleparate experimental studies giving it high confidence. Clustering

Immunology 51 (2012) 66–72

of all docked structures resulted in 1210 solutions (that differedin the placement of C1q relative to Fc by an Rmsdlig C1q of >5 Aafter best superposition with respect to Fc). Further filtering ofthe docking solutions involved both experimental data on puta-tive residues at the interface as well as the scoring of the dockinggeometries. A residue was counted as forming part of the interfaceif it was in contact with one or more residue of the partner protein.Two residues were considered in contact if any pair of coarse-grained pseudo-atoms of two residues was within a distance of 7 A.The docking scoring function in ATTRACT was shown to be quiteeffective in identifying near-native docking solutions especially incombination with force field weights on putative residues involvedin protein–protein interaction (Schneider and Zacharias, 2011).However, in order not to overlook a putative native-like solution,not only the best scoring docking solutions but also all solutionswith 25% of the top score were considered. Of these complexes asubset that fulfilled 80% of the experimental data on putative inter-face residues was selected for further refinement. This cutoff waschosen since test simulations indicated that the refinement proce-dure can still change the percentage of interface contacts by ∼20%such that the refinement procedure may result in perfect (100%)agreement with available experimental data.

It resulted in seven docked complexes that were refined fur-ther by MD simulations at all-atom level and included explicitsolvent during the final stage of refinement (see Section 2). Atthe final stage of refinement (in explicit solvent) resulting struc-tures could be assembled to three clusters of complexes whichkept more than 80% of the contacts that involved residues iden-tified by experiment to influence binding. The two smaller clustersof complexes (named model 2 and model 3) of these complexesshowed large conformational changes of the Fc antibody subdo-main (Fig. 3, backbone root-mean square deviation (Rmsd) in thebinding region > 7 A with respect to the Fc start structure). Thesechanges are significantly larger than what was observed in previouscomplexes with different protein partners (Ghai et al., 2007). Struc-tural model 1, however, agreed with all experimental mutagenesisdata and showed the least deviation of the partner proteins in thecomplex from the corresponding unbound conformations (overallRmsdbackbone < 1.5 A with respect to the corresponding experimen-tal crystal structures of Fc and C1q, respectively).

Model 1 also represented the highest populated cluster ofdocked complexes for which all members agreed to 80% or morewith the experimental data on interface residues, hence a represen-tative structure of this cluster serves as model structure for furtheranalysis. Interestingly, all three final models contact mainly one ofthe Fc monomers (chain A) that contain the ProA329 and ProA331

residues. Contacts to the second Fc monomer (chain B) involveresidues AspB270 and GluB272 of Fc (Fig. 3). Michaelson et al. (2006)found that mutagenesis of Pro329 to Ala (Pro329Ala) resulted in lossof complement activation if the mutation is introduced in both Fcmonomers but that activation is still possible if only one monomerhas been mutated. The model structure of the C1q–Fc complexexplains this observation since contacts predicted involve Pro329

and Pro331 of just one Fc monomer (chain A) and the same residuesin the other monomer (B) remain fully exposed to the solvent. Asindicated in Fig. 4 the structural model is also sterically fully com-patible with the presence of the carbohydrate structure connectedwith the Fc antibody part.

3.2. Identification of surface residues that mediate the interaction

The structural models of the C1q–Fc interaction allow the iden-

tification of possible contacts that mediate the interaction betweenthe two proteins which have not been considered in previousexperimental studies. This can be helpful for designing new muta-tions to prove or disprove a predicted contact, hence to validate the

S. Schneider, M. Zacharias / Molecular Immunology 51 (2012) 66–72 69

Fig. 2. (A) Cartoon representation of the globular C1q subunit (upper panel) and the IgG1–Fc subunit (lower panel). Several amino acid residues that have been foundexperimentally to contribute to the C1q–Fc interaction are labeled and indicated as van der Waals spheres (in case of C1q) or as stick model (for Fc). The cartoon colorcoding is blue for chain A, red for chain B and gray for chain C, respectively, of each subunit of the C1q trimer. In case of Fc chain A is in orange and chain B in yellow. (B)The electrostatic potential around C1q (upper panel) and Fc (lower panel) contoured at the 2kT level (k: Boltzmann constant and T: temperature: 300 K). Negative potentiali ew isp gure l

mits

Fsdt

s shown in red whereas regions of positive potential are indicated in blue. The viroposed interaction regions. (For interpretation of the references to color in this fi

odel, or can help to design mutations that may even enhance the

nteraction between the two proteins, which can help to improvehe efficiency of therapeutic antibodies. In contrast to mutagenesistudies that can help to identify putative interface residues that are

ig. 3. (A) Cartoon representation of the best scoring docking model that agreed with exolvent. Several residues important for binding are indicated as sticks. (B, C) Representativeviations of the Fc-structure from the start structure and agreed only partially with avaihe references to color in this figure legend, the reader is referred to the web version of th

approximately the same as in (A) indicating electrostatic complementarity of theegend, the reader is referred to the web version of this article.)

important for C1q–Fc interactions the final model of the C1q–Fc-

complex helps to assign one or more interaction partner residuesto each residue implicated in mediating C1q–Fc interactions. Onthe C1q side several mostly basic residues have been identified

perimental results on C1q–Fc interaction after refinement simulations in explicite structural models for two alternative C1q–Fc docking models that showed larger

lable experimental data. Color coding is the same as in Fig. 2. (For interpretation ofis article.)

70 S. Schneider, M. Zacharias / Molecular Immunology 51 (2012) 66–72

Fig. 4. Cartoon representation of C1q–Fc model complex structure in best agreement with available experimental data (same color coding as in Figs. 2 and 3). Interfaceresidues important for binding are indicated as sticks (Fc) or van der Waals spheres (C1q). The carbohydrate structures of the Fc molecule are indicated as gray sticks anda set of residues that were shown by mutagenesis experiments not to participate in binding are indicated as green sticks. Two proposed important interaction regions areh ues asc 329 an( rred to

toptacCrFslri2wabSti

et(iatr

tacts of Fc:ProA329 to residues ThrB100, ThrB102, IleB119, ValC162 andHisB117 of C1q. Substitution of the latter residue was shown to sig-nificantly reduce C1q–Fc binding affinity (Kojouharova et al., 2004).Residue Fc-ProA331 contacts C1q–ValC159 and ValC161. Interestingly,

Table 1Hydrogen bonding between C1q–Fc interface residues during refinement simulationof model 1.

C1q-trimer IgG1–Fc-dimer Occupancy

ArgC156 GluA333 100%ArgC182 GluA318 100%

ighlighted in the insets with C1q in surface representation and contacting Fc residlarity chain identifiers are only added in case of C1q residues. Residues Ser324, ProFor interpretation of the references to color in this figure legend, the reader is refe

o participate in the interaction with C1q head domain. In casef residues ArgB108, ArgB109, ArgB114, GluB162, ArgB129 and ArgC156

artner residues in hydrogen bonding or salt bridge forming dis-ance are observed in the final model structure (Fig. 4). This involvesn electrostatic contact between ArgB108 and Asp270 (on the B-hain of Fc). Substitution of residue Fc:Asp270 was shown to reduce1q–Fc binding (Idusogie et al., 2000; Thommesen et al., 2000). Foresidue Fc:LysA322 (chain A) contacts to GluC187 and Glu352 (samec chain) and also the nearby ArgC156 on C1q was found as a pos-ible network of contacting partners. However, this interaction isocated at the rim of the binding region and may depend on otheresidues nearby. It has been found that the involvement of Lys322

n C1q–Fc interaction depends on the subclass (Thommesen et al.,000). In the model residue ArgC156 of C1q can form a salt bridgeith Glu333 of Fc. However, in this configuration GluA333 is located

t the interface and is partially buried which may destabilize theinding. Indeed, it has been found that substitution of Glu333 byer (or Ala) increases the affinity of the complex presumably due tohe removal of an unfavorable buried charge in the protein–proteinnterface (Idusogie et al., 2001).

Substitution of Ser267 by a Glu was found to significantlynhance the C1q–Fc interaction (Moore et al., 2010). This result fitso the proposed model very well. In the model structure SerB267

chain B of Fc) is located near LysB123 and ArgB108 of C1q and

ntroducing a Glu at 267 strongly increases the electrostatic inter-ction with these two basic residues (see Fig. 4). Since most ofhe electrostatic contacts are mediated by residues at the rimegion of the C1q–Fc complex interface, we monitored the stability

stick model. Arrows indicate key residues for the protein–protein interaction. Ford Pro331 are in chain A of Fc whereas His268 and Asp270 are located in chain B of Fc.

the web version of this article.)

of these interactions during the MD simulations in explicit sol-vent. Many of the electrostatic interactions of the rim region formtransient contacts during the simulations with hydrogen bonds inrapid exchange with solvent. However, for some contacts hydrogenbonding with long life times was observed as indicated in Table 1.

The structural model also gives an explanation for the impor-tance of the two Pro329 and Pro331 residues for the interaction withC1q. In the model these two residues fit into a pocket that is locatedat the interface between chain B and C of C1q (Fig. 4). An overviewon the observed contacts in the model of the complex is given inTable 2. In contrast to most interactions with C1q and the chain Bof Fc, the interactions of the two Pro residues (located in chain A ofFc) with C1q involves several non-polar contacts. It predicts con-

ArgB161 GluB294 100%ThrB134 AsnB297 80%AsnB194 HisB268 50%ThrB112 GluB269 50%

S. Schneider, M. Zacharias / Molecular

Table 2Predicted residue–residue contactsa at the interface between C1q and IgG1–Fc.

C1q trimer IgG1–Fc dimer

ArgB108 AspB270

ArgB109 GluB269

ArgB114 GluB233

HisB117, ValC162, ThrB100, ThrB102, ValC179 ProA329

ArgB129 AspB265

ArgB161 GluB294

ArgC156 GluA333

GluC187 LysA322

ValC161, IleB119 ProA331

ValC159 SerA324

MetB158, PheB196 HisB268

LysB132, ArgB108 SerB267

o

iTs(C

osfotciPiiC

oT2fip

Fo

a A contact is defined if the distance between two heavy atoms belonging to pairf residues is <5 A in the final energy minimized structural model of the complex.

t shares the contact with ValC159 with SerA324 of the Fc structure.his contact offers an explanation for the observation that the sub-titution Ser324 by a slightly larger but more hydrophobic residueThr) with an additional methyl group increases the affinity of the1q–Fc complex (Moore et al., 2010).

In addition to explaining the importance of Pro329, Pro331 andf the Ser324Thr substitution for C1q–Fc interaction, the model alsouggests a hydrophobic contact between both partners at the inter-ace between chain B of C1q and chain B of Fc which involves HisB268

f the Fc molecule. It has been found by Moore et al. (2010) thathe substitution His268Phe actually increases the affinity of theomplex. In the present model residue HisB268 (of Fc) is locatedn a binding pocket formed by the hydrophobic residues (of C1q)heB196, MetB158 and AlaB164 (Fig. 4, Table 2). Residue PheB196 isn close proximity such that it probably could form efficient stack-ng interaction with a Phe268 which could explain the increase in1q–Fc binding observed experimentally.

Several residue substitutions have been found on the surfacef the IgG1 Fc antibody fragment that did not influence binding.his includes residues Lys276, Tyr 278 an Asp280 (Thommesen et al.,

000) as well as Val282, Val284 and His285 (Moore et al., 2010). In thenal structural model all these residues are outside of the predictedrotein–protein interface (indicated in Fig. 4).

ig. 5. Arrangement of C1q bound to the Fc domain of an IgG antibody in the contextf the full antibody structure bound to a bacterial surface.

Immunology 51 (2012) 66–72 71

The placement of the C1q in complex with the Fc-antibody seg-ment predicts a location on the opposite side of the Fab subunitsfor the arrangement of the rest of the complement C1 compo-nent (illustrated in Fig. 5). This corresponds to a sterically idealarrangement for a productive and simultaneous interaction of anti-bodies with an antigenic surface and with the C1 complement(Fig. 5). The model does not exclude the possibility of additionalcontacts between C1q and the Fab fragments themselves or thelinker between Fab fragments and Fc.

4. Conclusions

The atomic resolution docking structure of the C1q globularhead domain of the complement C1 factor with the IgG1 Fc domainprovides a working model for rationalizing available biochemicaldata on this important interaction. It explains experimental find-ings based on mutagenesis of putative interface residues and alsothe fact that only one Pro329/Pro331 motif is sufficient for C1q–Fcinteraction. Due to the direct assignment of contacting residueswithin the C1q–Fc interface of the structural model it offers thepossibility to specifically substitute interface residues in order toconfirm or to disprove suggested contacts. It is also sterically fullycompatible with the carbohydrate moiety of the Fc fragment andthe arrangement of the Fab fragments. In addition to explainingavailable mutagenesis data, the structural model suggests severalintramolecular contacts that could form the basis for the design ofnew Fc variants with a greater capacity to activate the complementsystem for example on binding to cancer cells or other target struc-tures. This includes not only residues in the vicinity of the importantPro329/Pro331 motif but also predicts hydrophobic regions on C1qthat interact with residue His268 and surrounding amino acids aswell as the region around residue Glu333 which is partially buried atthe interface. The globular head domain of C1q interacts also withthe Fc regions of other antibody types (e.g. IgM). It will be of interestto use the same docking and refinement approach for generatingstructural models of these interactions in the future.

Acknowledgements

We thank Drs C. Beier, S. de Vries and P. Setny for helpful dis-cussions. We thank the Deutsche Forschungsgemeinschaft (DFG)for financial support (grant Za-153/5-3).

References

Burton, D.R., Boyd, J., Brampton, A.D., Easterbrook-Smith, S.B., Emanuel, E.J., Novotny,J., Rademacher, T.W., van Schravendijk, M.R., Sternberg, M.J.E., Dwek, R.A., 1980.The C1q receptor site on immunoglobin G. Nature 288, 338–344.

Case, D.A., Darden, T.A., Cheatham III, T.E., Simmerling, C.L., Wang, J., Duke, R.E., Luo,R., Walker, R.C., Zhang, W., Merz, K.M., Roberts, B.P., Wang, B., Hayik, S., Roitberg,A., Seabra, G., Kolossvai, I., Wong, K.F., Paesani, F., Vanicek, J., Liu, J., Wu, X.,Brozell, S.R., Steinbrecher, T., Gohlke, H., Cai, Q., Ye, X., Wang, J., Hsieh, M.-J., Cui,G., Roe, D.R., Mathews, D.H., Seetin, M.G., Sagui, C., Babin, V., Luchko, T., Gusarov,S., Kovalenko, A., Kollman, P.A., 2010. AMBER 11. University of California, SanFrancisco.

Duncan, A.R., Winter, G., 1988. The binding site for C1q on IgG. Nature 332, 738–740.Ehrnthaller., C., Ignatius, A., Gebhard, F., Huber-Lang, M., 2011. New insights of an old

defense system: structure, function, and clinical relevance of the complementsystem. Mol. Med. 17, 317–329.

Fiorucci, S., Zacharias, M., 2010. Binding site prediction and improved scoring duringflexible protein–protein docking with ATTRACT. Proteins 78, 3131–3139.

Gaboriaud, C., Thielens, N.M., Gregory, L.A., Rossi, V., Fontecilla-Camps, J.C., Arlaud,G.J., 2004. Structure and activation of the C1 complex of complement: unravelingthe puzzle. Trends Immunol. 25, 368–373.

Gaboriaud, C., Juanhuix, J., Gruez, A., Lacroix, M., Darnault, C., Pignol, D., Verger, D.,Fontecilla-Camps, J.C., Arlaud, G.J., 2003. The crystal structure of the globular

head of complement protein C1q provides a basis for its versatile recognitionproperties. J. Biol. Chem. 278, 46974–46982.

Gadjeva, M.G., Rouseva, M.M., Zlatarova, A.S., Reid, K.B., Kishore, U., Kojouharova,M.S., 2008. Interaction of human C1q with IgG and IgM: revisited. Biochemistry47, 13093–13102.

7 ecular

G

H

I

I

K

K

K

K

K

L

L

M

M

2 S. Schneider, M. Zacharias / Mol

hai, R., Waters, P., Roumenina, L.T., Gadjeva, M., Kojouharova, M.S., Reid, K.B.,Sim, R.B., Kishore, U., 2007. C1q and its growing family. Immunobiology 212,253–266.

insen, K., 1998. Analysis of domain motions by approximate normal mode calcu-lations. Proteins 33, 417–429.

dusogie, E.E., Presta, L.G., Gazzano-Santoro, H., Totpal, K., Wong, P.Y., Ultsch, M.,Meng, Y.G., Mulkerrin, M.G., 2000. Mapping of the C1q binding site on rituxan,a chimeric antibody with a human IgG1 Fc. J. Immunol. 164, 4178–4184.

dusogie, E.E., Wong, P.Y., Presta, L.G., Gazzano-Santoro, H., Totpal, K., Ultsch, M.,Mulkerrin, M.G., 2001. Engineered antibodies with increased activity to recruitcomplement. J. Immunol. 166, 2571–2575.

aul, M., Loos, M., 1997. Dissection of C1q capability of interacting with IgG. J. Biol.Chem. 272, 33234–33244.

ishore, U., Kojouharova, M.S., Reid, K.B., 2002. Recent progress in the understand-ing of the structure–function relationships of the globular head regions of C1q.Immunobiology 205, 355–364.

ishore, U., Ghai, R., Greenhough, T.J., Shrive, A.K., Bonifati, D.M., Gadjeva, M.G.,Waters, P., Kojouharova, M.S., Chakraborty, T., Agrawal, A., 2004. Structuraland functional anatomy of the globular domain of complement protein C1q.Immunol. Lett. 95, 113–128.

ojouharova, M.S., Tsacheva, I.G., Tchorbadjieva, M.I., Reid, K.B., Kishore, U., 2003.Localization of ligand-binding sites on human C1q globular head region usingrecombinant globular head fragments and single-chain antibodies. Biochim.Biophys. Acta 1652, 64–74.

ojouharova, M.S., Gadjeva, M.G., Tsacheva, I.G., Zlatarova, A., Roumenina, L.T.,Tchorbadjieva, M.I., Atanasov, B.P., Waters, P., Urban, B.C., Sim, R.B., Reid,K.B., Kishore, U., 2004. Mutational analyses of the recombinant globularregions of human C1q A, B, and C chains suggest an essential role for argi-nine and histidine residues in the C1q–IgG interaction. J. Immunol. 172,4351–4358.

ensink, M.F., Wodak, S.J., 2010. Docking and scoring protein interactions: CAPRI2009. Proteins 78, 3073–3084.

u, J., The, B.K., Wang, Y., Tan, Y.S., Lai, M.C., Reid, K.B.M., 2008. The classical andregulatory functions of C1q in immunity and autoimmunity. Cell. Mol. Immunol.5, 9–21.

ay, A., Zacharias, M., 2005. Accounting for global protein deformability dur-

ing protein–protein and protein–ligand docking. Biochim. Biophys. Acta 1754,225–231.

ay, A., Zacharias, M., 2008. Energy minimization in low-frequency normal modes toefficiently allow for global flexibility during systematic protein–protein docking.Proteins 70, 794–809.

Immunology 51 (2012) 66–72

Michaelson, T.E., Thommesen, J.E., Ihle, O., Gregers, T.F., Sandin, R.H., Brekke, O.H.,Sandlie, I., 2006. A mutant human IgG molecule with only one C1q binding sitecan activate complement and induce lysis of target cells. Mol. Immunol. 36,129–138.

Moore, G.L., Chen, H., Karki, S., Lazar, G.A., 2010. Engineered Fc variant antibodieswith enhanced ability to recruit complement and mediate effector function.MABS 2, 181–189.

Oganesyan, V., Gao, C., Shirinian, L., Wu, H., Dall’Acqua, W.F., 2008. Strural charac-terization of a human Fc fragment engineered for lack of effector functions. ActaCrystallogr. D64, 700–704.

Presta, L.G., 2008. Molecular engineering and design of therapeutic antibodies. Curr.Opin. Immunol. 20, 460–470.

Radaev, S., Motyka, S., Fridman, W.H., Sautes-Fridman, C., Sun, P.D., 2001. The struc-ture of a human type III Fc gamma receptor in complex with Fc. J. Biol. Chem.276, 16469–16477.

Roumenina, L.T., Ruseva, M.M., Zlatarova, A., Ghai, R., Kolev, M., Olova, N., Gadjeva,M., Agrawal, A., Bottazzi, B., Mantovani, A., Reid, K.B., Kishore, U., Kojouharova,M.S., 2006. Interaction of C1q with IgG1, C-reactive protein and pentraxin 3:mutational studies using recombinant globular head modules of human C1q A,B, and C chains. Biochemistry 45, 4093–4104.

Saphire, E.O., Parren, P.W., Pantophlet, R., Zwick, M.B., Morris, G.M., Rudd, P.M.,Dwek, R.A., Stanfield, R.L., Burton, D.R., Wilson, I.A., 2001. Crystal structure ofa neutralizing human IGG against HIV-1: a template for vaccine design. Science293, 1155–1159.

Sarma, J.V., Ward, P.A., 2011. The complement system. Cell Tissue Res. 343, 227–235.Schneider, S., Zacharias, M., 2011. Optimizing the scoring during unbound

protein–protein docking including protein binding site predictions. J. Mol.Recognit. 25, 15–23.

Thommesen, J.E., Michaelson, T.E., Loset, G.A., Sandlie, I., Brekke, O.H., 2000. Lysine322 in the human IgG3 CH2 domain is crucial for antibody dependent comple-ment activation. Mol. Immunol. 36, 129–138.

Trouw, L.A., Daha, M.R., 2011. Role of complement in innate immunity and hostdefense. Immunol. Lett. 138, 35–37.

Zacharias, M., 2003. Protein–protein docking using a reduced model. Protein Sci. 12,1271–1282.

Zacharias, M., 2010. Accounting for conformational changes during protein–protein

docking. Curr. Opin. Struct. Biol. 16, 194–200.

Zlatarova, A.S., Rouseva, M., Roumenina, L.T., Gadjeva, M., Kolev, M., Dobrev, I., Olova,N., Ghai, R., Jensenius, J.C., Reid, K.B., Kishore, U., Kojouharova, M.S., 2006. Exis-tence of different but overlapping IgG- and IgM-binding sites on the globulardomain of human C1q. Biochemistry 45, 9979–9988.