MC

Ta

Sb

c

a

ARRAA

KCCIDI

1

scrsfl2ctpsacEmca

A

0d

Molecular Immunology 47 (2009) 332–339

Contents lists available at ScienceDirect

Molecular Immunology

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

odelling of the membrane receptor CXCR3 and its complexes with CXCL9,XCL10 and CXCL11 chemokines: Putative target for new drug design

iziana Trottaa,c,1, Susan Costantinia,b,∗,1, Giovanni Colonnaa

Department of Biochemistry and Biophysics and CRISCEB - (Interdepartmental Research Center for Computational and Biotechnological Sciences),econd University of Naples, via Costantinopoli 16, 80138 Naples, ItalyCROM (Oncology Research Centre of Mercogliano) “Fiorentino Lo Vuolo”, via Ammiraglio Bianco, 83013 Mercogliano, ItalyDoctorate in Computational Biology - CRISCEB - Second University of Naples, Italy

r t i c l e i n f o

rticle history:eceived 30 July 2009eceived in revised form 31 August 2009ccepted 3 September 2009vailable online 1 October 2009

a b s t r a c t

The chemokines play a key role in immune and inflammatory responses by promoting recruitment andactivation of different subpopulations of leukocytes. These comprise over 50 proteins grouped into fourclasses, in basis to the arrangement of conserved cysteine residues within the sequence.

CXCL9, CXCL10 and CXCL11 are the members of the family of ELR − CXC chemokines and bind thesame CXCR3 receptor. During the past few years, several studies have demonstrated a pathogenetic roleof CXCR3 and its ligands in many human inflammatory diseases. The blockade of CXCR3 interactions with

eywords:hemokineshemokine receptors

mmune responserug design

nflammation

its ligands has been suggested as a possible therapeutic target for the treatment of these diseases.Therefore, we modelled the three-dimensional structure of CXCL9 and CXCR3, and, successively, of the

CXCL9/CXCR3 complex in comparison to CXCL10/CXCR3 and CXCL11/CXCR3 complexes. We have thenshown the structural determinants of these interactions and their physico-chemical features. Finally, theinteraction residues involved in the formation of the complexes have been highlighted and analyzed in

desig

order to be used for drug

. Introduction

CXCL9, CXCL10 and CXCL11 are members of a family ofmall (8–10 kDa) proteins, the chemokines (or chemoattractantytokines). They play a key role in immune and inflammatoryesponses by promoting recruitment and activation of differentubpopulations of leukocytes, hence they have important proin-ammatory and immune modulatory functions (Booth et al.,002). There are over 50 known chemokines grouped into fourlasses, designated XC, CC, CXC and CX3C chemokines, based onhe arrangement of conserved cysteine (C) residues within therotein sequence (Robert, 2005). CXC chemokines have four con-erved cysteines and are distinguished by the presence of onemino acid between the first and second cysteine. These CXChemokines can be further divided into two groups (ELR+ and

LR−) according to the presence or absence of the tripeptideotif glutamic acid–leucine–arginine (ELR) N-terminal to the first

ysteine residue (Rotondi et al., 2007). Some chemokines medi-te specific aspects of the neoplastic phenotype, including cell

∗ Corresponding author at: CROM “Fiorentino Lo Vuolo”, via Ammiraglio Bianco,vellino, 83013 Mercogliano, Italy. Tel.: +39 0825 1911730; fax: +39 0825 1911705.

E-mail address: susan.costantini@unina2.it (S. Costantini).1 These authors contributed equally to this work.

161-5890/$ – see front matter © 2009 Elsevier Ltd. All rights reserved.oi:10.1016/j.molimm.2009.09.013

n.© 2009 Elsevier Ltd. All rights reserved.

growth, decreased sensitivity to cell death, angiogenesis, inva-sion and metastasis of cancer cells, whereas others are involvedin the antitumor immune response (Robert, 2005). Interestingly, asshown by site-directed mutagenesis studies, the presence or theabsence of an ELR motif in the chemokines-amino acid sequenceseems to correlate with their angiogenic or angiostatic activ-ity, respectively. Thus, ELR + CXC chemokines have been linkedto angiogenesis, whereas the ELR − CXC chemokines, includinginterferon (IFN)-� inducible protein 10 (IP-10/CXCL10), monokineinduced by IFN-� (MIG/CXCL9), and IFN-inducible T cell � chemoat-tractant (I-TAC/CXCL11), antagonize angiogenesis and appear tobe important in HCV pathogenesis (Rotondi et al., 2007; Zeremskiet al., 2008). The transmission of chemokine-encoded messages ismediated by specific cell-surface G protein-coupled receptors withseven transmembrane domains (Rotondi et al., 2007). The recep-tors follow a similar nomenclature to the chemokines, with CCreceptors interacting with CC chemokines and CXC receptors withCXC chemokines (Booth et al., 2004). Chemokine receptors com-prise 10 CCR family members, 7 CXCR family members and otherreceptors (XCR1, CCRL1 and 2, and CX3CR1). The chemokine sys-

tem also includes at least 3 “silent” receptors that bind ligands withhigh affinity but do not elicit signal transduction. Some chemokinesbind to multiple receptors and some receptors in turn bind multi-ple chemokines, whereas certain chemokines interact with singlereceptor and some receptors bind only one chemokine (Ali and

Immun

Lar2aCaamarsC

ap1tctfa2frC

ritrpahms1fsOp(

t(oifhpfCdw�CmH(caetb

s

T. Trotta et al. / Molecular

azennec, 2007; Zlotnik and Yoshie, 2000). CXCL9 as well as CXCL10nd CXCL11 (Loetscher et al., 1998) binds and activates the sameeceptor CXCR3 (chemokine (C-X-C motif) receptor 3) (Booth et al.,004; Cole et al., 1998). CXCR3 is mainly expressed on activated Tnd natural killer (NK) cells (Zeremski et al., 2007). While CXCL11,XCL10, and CXCL9 are agonists for CXCR3, they can also act asntagonists for CCR3 (Loetscher et al., 2001). Binding competitionnd inhibition of the functional responses show that CXCL11 is theost potent antagonist followed by CXCL9 and CXCL10 (Loetscher

nd Clark-Lewis, 2001). It is interesting that the same potencyanking is observed for the agonistic activity mediated via CXCR3,uggesting the existence of binding-relevant homologies betweenXCR3 and CCR3.

Although these three ligands are all induced by IFN-�, theyppear to mediate distinct biological phenomena in vivo. They areroduced by macrophages as well as other cell types (Cole et al.,998; Laich et al., 1999; Tensen et al., 1999). This may be relatedo differential expression of these ligands as it has been seen inardiac and skin allograft rejection, atherosclerosis, host responseo infection, and inflammatory skin disease. Alternatively, the dif-erent biological outcomes may also be related to the differentialctivation of CXCR3 by CXCL10, CXCL9, and CXCL11 (Colvin et al.,004). However, CXCL11 also shows a number of functional dif-erences from CXCL10 and CXCL9, and it has significantly highereceptor binding affinity and is a more potent chemoattractant thanXCL10 or CXCL9 (Booth et al., 2004).

Tumor cells aberrantly express chemokines and/or chemokineeceptors, and the interaction of chemokine ligand–receptor pairs isncreasingly implicated as a mediator of tumor growth and metas-asis. In several malignancies, tumor cell expression of chemokineeceptors is associated with more aggressive disease and poorerrognosis (Cole et al., 1998; Zeremski et al., 2007; Loetscher etl., 2001; Loetscher and Clark-Lewis, 2001). In particular, CXCR3as now been identified in a variety of malignant cells, includingelanoma, breast and prostate carcinomas, neuroblastoma, and a

ubset of B cell lymphomas (O’Donovan et al., 1999; Cole et al.,998; Laich et al., 1999; Tensen et al., 1999; Colvin et al., 2004). Theunctional significance and clinical implications of CXCR3 expres-ion by tumor cells remain to be determined (Walser et al., 2006).ther study suggests that CXCL9 and CXCL11, in addition to CXCL10,lay a role in the accumulation of Th1 cells into sarcoid lungNishioka et al., 2007).

Moreover, CXCL9, CXCL10, and CXCL11, are powerful chemoat-ractants for lymphocytes. The tumor-infiltrating lymphocytesTILs) are known to play a role in the tumor–host reaction in vari-us types of neoplasms, and the degree of lymphocytic infiltrations known to be a significant determinant of treatment outcomeor patients with a variety of malignancies. These TILs includeelper and suppressor T lymphocytes, natural killer cells, B lym-hocytes, and macrophages. In solid tumors T lymphocytes wereound, whereas the number of B cells resulted proportionally lower.onsistent with the promise of T lymphocytes, some studies haveemonstrated that TILs in hepatocarcinoma (HCC) are associatedith a good prognosis. In detail, these studies suggest that IFN-is produced by TILs and induces chemokines such as CXCL9 and

XCL10 (Hirano et al., 2007). Therefore, these chemokines may pro-ote the recruitment of lymphocytes to HCC and release from theCC cells may induce lymphocyte infiltration. Ruehlmann et al.

2001) demonstrated that CXCL9 chemokine gene therapy, whenombined with IL-2, suppressed the growth and dissemination ofmurine colon carcinoma. Therefore, these suggested that the

xpression of CXCL9 and CXCL10 might lead to lymphocytic infil-ration into HCC, and gene therapy with these CXC chemokines maye effective for patients with HCC.

In this paper, at first, we simulated the three-dimensionaltructure of CXCL9 and CXCR3, and, successively, we modelled

ology 47 (2009) 332–339 333

the CXCL9/CXCR3 complex in comparison to CXCL10/CXCR3 andCXCL11/CXCR3 complexes. Then we have evaluated in detail theinteraction residues involved in the formation of the complexesand their properties as important structural features to be used fordrug design.

2. Methods

2.1. Modelling of both CXCL9 and CXCR3

The sequences of both CXCL9 and CXCR3 were analyzed withBLAST (Altschul et al., 1990) to find similar proteins in databases.The three-dimensional model of both CXCL9 and CXCR3 wereperformed according to the template-based modelling strategyusing the template structures of human CXCL10 and CXCL11 (PDB-code: 1LV9 and 1RJT, respectively) for CXCL9 (Booth et al., 2002,2004) and that of bovine rhodopsin (PDB code: 1F88) for CXCR3(Palczewski et al., 2000). Protein sequences were aligned withCLUSTALW (Thompson et al., 1994). Few manual refinements wereadded to account for the secondary position of the template andtarget protein sequences, as well as to avoid the presence of gapsand insertions in secondary structure elements, which are oftenresponsible for wrong models.

The MODELLER9v5 program was used to build 10 full-atommodels of both CXCL9 and CXCR3 (Sali and Blundell, 1993). Toselect the best models, we used the ProsaII program to check thefitness of the sequences relative to the obtained structures and toassign a scoring function (Sippl, 1993) and Procheck program toevaluate the stereochemical quality (Laskowski et al., 1993). Sec-ondary structures were assigned by the DSSP program (Kabsch andSander, 1983). Search for structural classification was performed onthe CATH database (Orengo et al., 1997). Secondary structure pre-dictions were performed with JPRED [29] server (Cuff and Barton,2000). The phosphorylation sites have been predicted for all pro-teins according to NetPhos Server (Blom et al., 1999).

2.2. Modelling of the complexes between CXCL9, CXCL10 andCXCL11 with CXCR3

CLUSPRO (Comeau et al., 2004) and PATCHDOCK (Schneidman-Duhovny et al., 2003) web servers were used to simulate theCXCL9/CXCR3 complex using the obtained models for CXCL9and CXCR3 by the template-based modelling strategy. CLUSPROresulted the best docking program in CAPRI 2005 experimentwith a success rate of about 71%. It is a fast algorithm for fil-tering docked conformations with good surface complementarity,and ranking them based on their clustering properties; the freeenergy filters select complexes with lowest desolvation and elec-trostatic energies. In particular, ZDOCK implemented in CLUSPROwas selected to perform the initial rigid-body docking, where scor-ing function includes a combination of shape complementarity.While, PatchDock is a geometry-based molecular docking algo-rithm aimed at finding docking transformations that yield goodmolecular shape complementarity. These complementary patchesare matched in order to generate candidate transformations andeach candidate transformation is further evaluated by a scoringfunction that considers both geometric fit and atomic desolvationenergy (Schneidman-Duhovny et al., 2005).

The CXCL10/CXCR3 and CXCL11/CXCR3 complexes were sim-ulated using the same procedure reported above and thecrystallographic structures of CXCL10 and CXCL11 and the CXCR3

model obtained by the template-based modelling strategy.

The selected complexes were minimized by using 500 steps ofenergy minimization under conjugate gradient algorithm in orderto optimize side chain conformations and avoid sterical clashesaccording to the commonly used procedure (Costantini et al., 2005,

334 T. Trotta et al. / Molecular Immunology 47 (2009) 332–339

F n boldi bels ib

2“atat(ebP

3

3

rmastaiafaZTisoiCrattmibCioCc�

�-adrenergic receptors for which the experimental models wereincomplete and the cartesian coordinates of some segments (i.e. N-terminal or extracellular loops) were lacking. The best model (Fig. 4)had Prosa Z-score of −3.47 and 96.2% of residues in most favored

ig. 1. Alignment of CXCL9, CXCL10 and CXCL11. Cysteine residues are evidenced in �-strands are reported in bold and those in the helix are evidenced in grey. The lay NetPhos server are underlined.

007, 2008a; Gianfrani et al., 2007; Paladino et al., 2008). TheProtein–Protein Interaction Server” (Jones and Thornton, 1996)nd NACCESS program (Hubbard et al., 1991) were used to iden-ify the amino acids at the interface in the CXCR3/CXCL9 complexnd to evaluate their solvent accessibility. The presence of puta-ive H-bonds and salt bridges were calculated with Hbplus programMcDonald and Thornton, 1994) and ESBRI web server (Costantinit al., 2008b), respectively. Moreover, the binding free energyetween the different chains was calculated by using the DCOM-LEX program (Liu et al., 2004).

. Results

.1. CXCL9 model

The three-dimensional model of CXCL9 (UniProt code: Q07325,egion: 23–91) was performed according to the template-basedodelling strategy using the template structures of human CXCL10

nd CXCL11 (PDBcode: 1LV9, and 1RJT) because the percentage ofequence identity between CXCL9 sequence and two template pro-eins resulted of 37% and 38%, respectively. In Fig. 1 we show thelignment of CXCL9, CXCL10 and CXCL11 sequences. Four cysteinesnvolved in two disulfide bridges in the template structures areligned to four related cysteines in the CXCL9 sequence. Startingrom this alignment with the reference structures, a set of 10 full-toms models was generated. The best model (Fig. 2) had Prosa-score of −4.53 and 88.5% of residues in most favored regions.hese values, compared with those of the template structures,ndicated that a good quality model has been created. Secondarytructure predictions made by JPRED programs agree with thebtained structure of CXCL9 by template-free modelling exceptn the N-terminal region. CXCL9 presents the typical structure ofXC and CC chemokines characterized by a short NH2-terminalegion, a large core, which is stabilized by the disulfide bondsnd is characterized by three antiparallel �-strands, and a COOH-erminal �-helix. The structure of the core is well-ordered, buthe NH2-terminal domain and the distal region of the COOH ter-

inus have high conformational flexibility. In particular, the loopn the area of N-terminal of CXCL9 is stabilized by two disulfideonds: Cys33–Cys58 and Cys31–Cys74. Moreover, the models ofXCL9, CXCL10 and CXCL11 were compared by structural super-

mposition and in terms of secondary structures. RMSD valuesbtained comparing the model of CXCL9 with both CXCL10 andXCL11 resulted of 0.609 and 1.108 Å, respectively. Moreover, theomparison of the secondary structures evidenced that helices and-strands are well conserved along the sequence, and few changes

. Secondary structure prediction made by JPred is reported for CXCL9. Amino acidsndicate N-terminal and N-loop. The most probable phosphorylation sites predicted

are observed in length. Certainly, these evaluations suggest thatthree CXC chemokine structures have a similar tertiary structureand the structural differences are located at the level of loops.

3.2. CXCR3 model

The three-dimensional model of CXCR3 (Uniprot code: P49682,region 18-334) was performed according to the comparative mod-elling strategy, using the template structures of bovine rhodopsin(PDB code: 1F88) (Palczewski et al., 2000). In Fig. 3 we showCXCR3/template alignment. We have selected the rhodopsin struc-ture as template for CXCR3 modelling because it had highersequence identity with CXCR3 sequence (22%) and the alignmentbetween CXCR3 and the rhodoposin presented the better E-value(10−9) respect to that obtained from the alignment between CXCR3and some �-adrenergic receptors ranged from 10−6 to 10−4. More-over we have selected the rhodopsin structure as template alsobecause the alignment between CXCR3 and rhodopsin sequencespresented a lower number of gaps respect to that with other

Fig. 2. 3D model of human CXCL9. The backbone ribbon is reported in green. The�-strands are indicated with yellow arrows and helices in red. The SS-bonds areevidenced. (For interpretation of the references to color in the citation of this figure,the reader is referred to the web version of the article.)

T. Trotta et al. / Molecular Immunology 47 (2009) 332–339 335

F n greyp

rscstpltttvimcAC

3C

tciao

ig. 3. Alignment of CXCR3 and rhodopsin. Amino acids in the helix are reported irediction made by JPred is reported for CXCR3.

egions. Also these values, compared with that of the templatetructure, indicated that our model had a good energetic and stereo-hemical quality as well as in the case of CXCL9 model. Secondarytructure predictions made by JPRED agree with the obtained struc-ure of CXCR3 by comparative modelling (see Fig. 3). CXCR3 modelresents seven transmembrane domains and three extracellular

oops (i.e. loops 1, 2 and 3), in agreement with experimental struc-ure for the bovine rhodopsin (Palczewski et al., 2000). Moreover,he models of CXCR3 and bovine rodopsin were compared by struc-ural superimposition and in terms of secondary structures. RMSDalue obtained comparing two models resulted of 2.21 Å. Compar-ng the secondary structures between CXCR3 and bovine rhodopsin

odels it is evident that the seven transmembrane helices are wellonserved, and few changes are observed in helix length (see Fig. 3).lso phosporylation sites were predicted in N-terminal, loop 3 and-terminal regions (see Supplementary Material).

.3. Modelling of CXCL9/CXCR3, CXCL10/CXCR3 andXCL11/CXCR3 complexes

The CXCL9/CXCR3 complex was obtained using the above men-

ioned CLUSPRO and PATCHDOCK web servers (Fig. 5). For thisomplex, we evaluated the interaction residues, the number ofnterchain H-bonds and of salt bridges, the interface surface areand binding energy (Table 1, Figs. 6 and 7). In the best modelbtained by PATCHDOCK the CXCL9 and CXCR3 chains form two

. The labels indicate seven helix regions and the loop regions. Secondary structure

H-bonds and one salt bridge. Numerous structure–activity studiesof chemokines have shown that these proteins possess two mainsites that interact with their receptors (Clark-Lewis et al., 1995;Loetscher and Clark-Lewis, 2001; Fernandez and Lolis, 2002), andhave led to a two-step model for the interaction, in which recep-tor binding and activation are dissociated. In the initial event, theN-loop region of the chemokine recognizes and binds the recep-tor (“docking”). This initial contact facilitates subsequent bindingand proper positioning of the flexible N-terminal region of thechemokine to the receptor, which leads to its activation (“trig-gering”). The region of the receptor responsible for chemokinedocking has been localized to N-terminal (Booth et al., 2002) (seeFig. 3). Our results are in agreement with the experimental data,which means that the N-loop region of CXCL9 interacts with theN-terminal region of CXCR3 and N-terminal of CXCL9 with theloops 2 and 3 of CXCR3 as suggested by Xanthou et al. (2003)(see Fig. 6).

Moreover, the complexes of CXCR3 with both CXCL10 andCXCL11 were obtained using the programs indicated in Section 2.Also for those complexes we evaluated the interaction residues,the number of interchain H-bonds and of salt bridges, the interface

surface area, and binding energy (Table 1, Figs. 6 and 7).

There are additional indications that the nature of CXCL11’sreceptor interaction is somewhat different from the other CXCR3binding chemokines. These data show that both the N-terminaland N-loop region contribute to the higher receptor binding affin-

336 T. Trotta et al. / Molecular Immunology 47 (2009) 332–339

Fig. 4. 3D model of human CXCR3. The backbone ribbon is reported in green butseven helices in red. (For interpretation of the references to color in the citation ofthis figure, the reader is referred to the web version of the article.)

Table 1Analysis of the three complexes in terms of interface surface area (Å2), number ofinteraction residues, interchain H-bonds and salt bridges evaluated for each chain.

Complexes Interfacesurfacearea

Number ofinteractionresidues

InterchainH-bonds

Saltbrigdes

CXCL9/CXCR3CXCR3 1026.89 27 2 1CXCL9 938.81 25 2 1

CXCL10/CXCR3CXCR3 1149.77 33 3 3CXCL10 1230.59 25 3 3

Fig. 5. 3D model of CXCL9/CXCR3 complex where CXCR3 is reported with green ribbon but(i.e. N-terminal, loop 1, loop 2 and loop 3 of the receptor are shown in red, yellow, blue aand grey, respectively). (For interpretation of the references to color in the citation of this

CXCL11/CXCR3CXCR3 1229.09 36 4 4CXCL11 1233.6 25 4 4

ity and activity of CXCL11 compared with CXCL10 and CXCL9 andsuggest that these two domains are the major functional determi-nants (Clark-Lewis et al., 2003). CXCL11 has the highest affinity forCXCR3, and is also the most potent agonist (Meyer et al., 2001;Sauty et al., 2001; Clark-Lewis et al., 2003). Deletion of the N-terminal three residues of CXCL11 resulted in a potent antagonistthat competes for binding of CXCR3 with the full-length protein.In contrast, deletion of the N-terminal three residues of CXCL10 orCXCL9 leads to loss of binding capacity (Clark-Lewis et al., 2003),indicating that the N-loop region alone is not sufficient for recep-tor binding. CXCL10/CXCL11 hybrids in which residues 12–17 inthe N-loop of CXCL10 are replaced with residues 12–17 of CXCL11show comparable binding to wild-type CXCL11, indicating thatmuch of the additional binding strength of CXCL11 originates inthis region of the chemokine (Clark-Lewis et al., 2003). Moreover,

it was demonstrated that the second extracellular loop of CXCR3 isessential for receptor activation in response to all CXCR3 ligands. Incontrast, the N-terminal and first extracellular loop of CXCR3 playsome role in CXCL11-mediated activation but are dispensable forCXCL9-induced signalling (Xanthou et al., 2003).

CXCL9 with cyan ribbon. In detail, the loops of the interaction regions are evidencednd magenta, respectively, and N-terminal and N-loop of the chemokine in orangefigure, the reader is referred to the web version of the article.)

T. Trotta et al. / Molecular Immunology 47 (2009) 332–339 337

Fig. 6. Interaction residues between CXCL9/CXCL10/CXCL11 and CXCR3 (a) an

Fig. 7. Binding free energies for the CXCL9/CXCR3, CXCL10/CXCR3 andCXCL11/CXCR3 complexes. The bars represent the binding energies (expressed inkcal/mol).

d between CXCR3 and CXCL9/CXCL10/CXCL11 are reported underlined.

Analysing the CXCL10/CXCR3 and CXCL11/CXCR3 complexes,the N-loop of both CXCL10 and CXCL11 interacts with the N-terminal region of CXCR3, and the loops 1, 2 and 3 of CXCR3 withthe related N-terminal region of each chemokine.

Finally, comparing three complexes of CXCR3 with CXCL9,CXCL10 and CXCL11, CXCR3 shows the highest affinity for CXCL11in terms of binding energy (see Fig. 7) and the highest value of inter-face surface area (Å2) and the highest number of H-bonds, of saltbridges and of interaction residues (Table 1) in according to pub-lished data (Meyer et al., 2001; Sauty et al., 2001; Clark-Lewis et al.,2003).

4. Discussion

During the past few years, several studies have demonstrateda pathogenenetic role of CXCR3 and its ligands in human inflam-

3 Immun

mowtIastcendiltawt

tsctwrdoTwaepii

Cirr(Loette

raN2asleoCmPfc

aphs

38 T. Trotta et al. / Molecular

atory diseases suggesting the involvement of various segmentsf their sequences. Therefore, the blockade of CXCR3 interactionsith its ligands in vivo has been suggested as a possible therapeu-

ic goal for the treatment of these disorders (Xanthou et al., 2003).n this paper we show the structural determinants of these inter-ctions, their physico-chemical features, and a close view of thepecific residues involved. Chemokines are tempting therapeuticargets for a wide range of diseases, because they are specifi-ally upregulated during inflammation. CXCR3 have also effects onndothelial cells involved in angiogenesis or angiostasis. Unfortu-ately all the underlying molecular mechanisms of the above citediseases, however, remain strongly elusive. Moreover some stud-

es have suggested that the expression of CXCL9 and CXCL10 mightead to lymphocytic infiltration into hepatocarcinoma, and the geneherapy with these CXC chemokines may be effective for patientsffected from this cancer as well as CXCL9 chemokine gene therapy,hen combined with IL-2, suppressed the growth and dissemina-

ion of a murine colon carcinoma (Ruehlmann et al., 2001).One intriguing question related to the chemokine family and

heir receptors is the apparent redundancy of the chemokineystem, with most receptors being activated by more than onehemokine and many chemokines binding to more than one recep-or. Approximately 50 chemokines have been found to interactith about 16 GPCR chemokine receptors, implying that there is

edundancy in the chemokine system. Some studies in vivo haveemonstrated that chemokines that activate the same receptorften have unique biological functions in vivo (Proudfoot, 2002).his may be related to differential chemokine expression in vivo asell as differential receptor activation by different chemokine lig-

nds. In this study, we have explored the structural basis for theseffects in the CXCR3 receptor ligand system because our aim is torovide a detailed examination of the interaction residues involved

n the formation of the complexes, and, in particular, those of CXCR3n order to use for new drug design.

We have modelled and analyzed three CXCL9/CXCR3,XCL10/CXCR3 and CXCL11/CXCR3 complexes. Three result-

ng chemokines always interact with their receptor by N-terminalegion and N-loop but they interact with receptor by N-terminalegion and three extracellular loops according to precedent studiesXanthou et al., 2003; Clark-Lewis et al., 1995; Loetscher and Clark-ewis, 2001; Fernandez and Lolis, 2002). Moreover the analysis ofur complexes shows that the N-loop of all three chemokines isssential for binding the N-terminal region of CXCR3 in agreemento Clark-Lewis et al. (2003) whereas the loop 1 of CXCR3 is essentialo bind only CXCL11 and CXCL10 as well as indicated by Xanthout al. (2003).

In detail, we analyzed also the physical–chemical properties ofesidues present in these regions in CXCR3: (i) N-terminal, loop 1nd loop 2 contain some aromatic residues (Phe, Tyr and Trp); (ii)-terminal presents three negatively charged residues (3 Glu), loopone (Asp) but loop 3 three (2 Asp and 1 Glu) and (iii) both loop 2

nd loop 3 have two positively charged residues (2 Arg). These datauggest that the predominant interaction between CXCR3 and itsigands is on electrostatic basis and is also favored from the pres-nce of positively charged residues located in N-terminal regionf three chemokines (i.e. three in CXCL9 and CXCL11 and two inXCL10). Moreover, the presence of aromatic residues stabilizeainly the interaction between CXCR3 and CXCL11, having two

he residues in N-terminal and might play an important role toavour the stacking interactions with putative drugs and organicompounds.

Another intriguing point is the presence of all these proteinst the phosphorylation sites. To assess the effects of the phos-horylation on human CXCR3 as well as on its three ligands weave analyzed the structures to find the most probable putativeites of phosphorylation. The presence of phosphorylation sites is

ology 47 (2009) 332–339

important to assess the molecular mechanisms trough which theseproteins and/or their complexes regulate the biological function.

Our data indicate that in CXCR3 the phosphorylation sitesare mostly located in N-terminal and C-terminal segments (seeSupplementary Material). Those located in the N-terminus (Ser 8,9, 19, 69 and Tyr 12) have a high likelihood to be involved in bind-ing. Even the locations of phosphorylation sites of CXCL9, CXCL10and CXCL11 are located in the binding regions of these molecules.This suggests that these residues can be involved in some way inthe control of protein–protein binding and, as a consequence, in thecontrol of the CXCR3 function.

The three CXCR3/CXCLX complexes highlights that the inter-action between the proteins occurs primarily through theinvolvement of the N-terminal segment and loops 2 and 3 of CXCR3and that thus these regions can be involved in regulating the bind-ing. However overall these results could be helpful to identify themost relevant inflammatory pathways present in CXCR3 relateddiseases and in future will be confirmed also by experimental stud-ies because they could be influenced by the accuracy of the obtainedmodels.

Appendix A. Supplementary data

Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.molimm.2009.09.013.

References

Ali, S., Lazennec, G., 2007. Chemokines: novel targets for breast cancer metastasis.Cancer Metastasis Rev. 26, 401–420.

Altschul, S.F., Gish, W., Miller, W., Myers, E., Lipman, D.J., 1990. Best local alignmentsearch tool. J. Mol. Biol. 215, 403–410.

Blom, N., Gammeltoft, S., Brunak, S., 1999. Sequence- and structure-based predictionof eukaryotic protein phosphorylation sites. J. Mol. Biol. 294, 1351–1362.

Booth, V., Keizer, D.W., Kamphuis, M.B., Clark-Lewis, I., Sykes, B.D., 2002. The CXCR3binding chemokine IP-10/CXCL10: structure and receptor interactions. Bio-chemistry 41, 10418–10425.

Booth, V., Clark-Lewis, I., Sykes, B.D., 2004. NMR structure of CXCR3 bindingchemokine CXCL11 (ITAC). Protein Sci. 13, 2022–2028.

Clark-Lewis, I., Kim, K.S., Rajarathnam, K., Gong, J.H., Dewald, B., Moser, B., Baggiolini,M., Sykes, B.D., 1995. Structure–activity relationships of chemokines. J. Leukoc.Biol. 57, 703–711.

Clark-Lewis, I., Mattioli, I., Gong, J.H., Loetscher, P., 2003. Structure–function rela-tionship between the human chemokine receptor CXCR3 and its ligands. J. Biol.Chem. 278, 289–295.

Cole, K.E., Strick, C.A., Paradis, T.J., Ogborne, K.T., Loetscher, M., Gladue, R.P., Lin,W., Boyd, J.G., Moser, B., Wood, D.E., Sahagan, B.G., Neote, K., 1998. Interferon-inducible T cell � chemoattractant (I-TAC): a novel ELR-CXC chemokine withpotent activity on activated T cells through selective high affinity binding toCXCR3. J. Exp. Med. 187, 2009–2021.

Colvin, R.A., Campanella, G.S.V., Sun, J., Luster, A.D., 2004. Intracellular domain ofCXCR3 that mediate CXCL9, CXCL10, and CXCL11 function. J. Biol. Chem. 279,30219–30227.

Comeau, S.R., Gatchell, D.W., Vajda, S., Camacho, C.J., 2004. ClusPro: an automateddocking and discrimination method for the prediction of protein complexes.Bioinformatics 20, 45–50.

Costantini, S., Rossi, M., Colonna, G., Facchiano, A.M., 2005. Modelling of HLA-DQ2and its interaction with gluten peptides to explain molecular recognition inceliac disease. J. Mol. Graph. Model. 23, 419–431.

Costantini, S., Colonna, G., Facchiano, A.M., 2007. Simulation of conformationalchanges occurring when a protein interacts with its receptor. Comput. Biol.Chem. 31, 196–206.

Costantini, S., Buonocore, F., Facchiano, A.M., 2008a. Molecular modelling of co-receptor CD8�� and its complex with MHC class I and T-cell receptor in seabream (Sparus aurata). Fish Shellfish Immunol. 25, 782–790.

Costantini, S., Colonna, G., Facchiano, A.M., 2008b. ESBRI: a software for evaluatingthe salt bridges in proteins. Bioinformation 3, 137–139.

Cuff, J.A., Barton, G.J., 2000. Application of enhanced multiple sequence alignmentprofiles to improve protein secondary structure prediction. Proteins: Struct.Funct. Genet. Proteins 40, 502–511.

Fernandez, E.J., Lolis, E., 2002. Structure, function, and inhibition of chemokines.Annu. Rev. Pharmacol. Toxicol. 42, 469–499.

Gianfrani, C., Siciliano, R., Facchiano, A.M., Camarca, A., Mazzeo, F.M., Costantini,S., Savati, V.M., Maurano, F., Mazzarella, G., Iaquinto, G., Bergamo, P., Rossi, M.,2007. Transamidation of wheat flour inhibits the response to gliadin of intestinalT cells in celiac disease. Gastroenterology 133, 780–789.

Immun

H

H

J

K

L

L

L

L

L

L

M

M

N

O

O

P

P

T. Trotta et al. / Molecular

irano, S., Iwashita, Y., Sasaki, A., Kai, S., Ohta, M., Kitano, S., 2007. Increased mRNAexpression of chemokines in hepatocellular carcinoma with tumor-infiltratinglymphocytes. J. Gastroenterol. Hepatol. 22, 690–696.

ubbard, S.J., Campbell, S.F., Thornton, J.M., 1991. Molecular recognition. Confor-mational analysis of limited proteolytic sites and serine proteinase proteininhibitors. J. Mol. Biol. 220, 507–530.

ones, S., Thornton, J.M., 1996. Principles of protein–protein interactions derivedfrom structural studies. PNAS 93, 13–20.

absch, W., Sander, C., 1983. Dictionary of protein secondary structure: patternrecognition of hydrogen-bonded and geometrical features. Biopolymers 22,2577–2637.

aich, A., Meyer, M., Werner, E.R., Werner-Felmayer, G., 1999. Structure and expres-sion of the human small cytokine B subfamily member 11 (SCYB11/formerlySCYB9B, alias I-TAC) gene cloned from IFN-�-treated human monocytes (THP-1).J. Interferon Cytokine Res. 19, 505–513.

askowski, R.A., MacArthur, M.W., Moss, D.S., Thornton, J.M., 1993. PROCHECK—aprogram to check the stereochemical quality of protein structures. J. Appl. Cryst.26, 283–291.

iu, S., Zhang, C., Zhou, H., Zhou, Y., 2004. A physical reference state unifies thestructure-derived potential of mean force for protein folding and binding. Pro-teins 56, pp. 93e101.

oetscher, M., Loetscher, P., Brass, N., Meese, E., Moser, B., 1998. Lymphocyte-specificchemokine receptor CXCR3: regulation, chemokine binding and gene localiza-tion. Eur. J. Immunol. 28, 3696–3705.

oetscher, P., Pellegrino, A., Gong, J.H., Mattioli, I., Loetscher, M., Bardi, G., Baggiolini,M., Clark-Lewis, I., 2001. The ligands of CXC chemokine receptor 3, I-TAC, MIG,and IP-10, are natural antagonists for CCR3. J. Biol. Chem. 276, 2986–2991.

oetscher, P., Clark-Lewis, I., 2001. Agonistic and antagonistic activities ofchemokines. J. Leukoc. Biol. 69, 881–884.

cDonald, I.K., Thornton, J.M., 1994. Satisfying hydrogen bonding potential in pro-teins. J. Mol. Biol. 238, 777–793.

eyer, M., Hensbergen, P.J., van der Raaij-Helmer, E.M., Brandacher, G., Margreiter,R., Heufler, C., Koch, F., Narumi, S., Werner, E.R., Colvin, R., Luster, A.D., Tensen,C.P., Werner-Felmayer, G., 2001. Cross reactivity of three T cell attracting murinechemokines stimulating the CXC chemokine receptor CXCR3 and their inductionin cultured cells and during allograft rejection. Eur. J. Immunol. 31, 2521–2527.

ishioka, Y., Manabe, K., Kishi, J., Wang, W., Inayama, M., Azuma, M., Sone, S.,2007. CXCL9 and 11 in patients with pulmonary sarcoidosis: a role of alveolarmacrophages. Clin. Exp. Immunol. 149, 317–326.

’Donovan, N., Galvin, M., Morgan, J.G., 1999. Physical mapping of the CXCchemokine locus on human chromosome 4. Cytogenet. Cell. Genet. 84, 39–42.

rengo, C.A., Michie, A.D., Jones, S., Jones, D.T., Swindells, M.B., Thornton, J.M.,1997. CATH—a hierarchic classification of protein domain structures. Structure

5, 1093–1108.

aladino, A., Costantini, S., Colonna, G., Facchiano, A.M., 2008. Molecular modellingof miraculin: structural analyses and functional hypotheses. BBRC 367, 26–32.

alczewski, K., Kumasaka, T., Hori, T., Behnke, C.A., Motoshima, H., Fox, B.A., Le Trong,I., Teller, D.C., Okada, T., Stenkamp, R.E., Yamamoto, M., Miyano, M., 2000. Crystalstructure of rhodopsin: a G protein-coupled receptor. Science 289, 739–745.

ology 47 (2009) 332–339 339

Proudfoot, A.E., 2002. Chemokine receptors: multifaceted therapeutic targets. Nat.Rev. Immunol. 2, 106–115.

Robert, L.R., 2005. Chemokine as attractive targets in liver carcinogenesis. Am. J.Gastroenterol. 100, 499–501.

Rotondi, M., Chiovato, L., Romagnani, S., Serio, M., Romagnani, P., 2007. Role ofchemokines in endocrine autoimmune disease. Endocr. Soc. 28, 492–520.

Ruehlmann, J.M., Xiang, R., Niethammer, A.G., Ba, Y., Pertl, U., Dolman, C.S., Gillies,S.D., Reisfeld, R.A., 2001. MIG (CXCL9) chemokine gene therapy combines withantibody-cytokine fusion protein to suppress growth and dissemination ofmurine colon carcinoma. Cancer Res. 61, 8498–8503.

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

Sauty, A., Colvin, R.A., Wagner, L., Rochat, S., Spertini, F., Luster, A.D., 2001. CXCR3internalization following T cell-endothelial cell contact: preferential role ofIFN-inducible T cell � chemoattractant (CXCL11). J. Immunol. 167, 7084–7093.

Schneidman-Duhovny, D., Inbar, Y., Polak, V., Shatsky, M., Halperin, I., Benyamini,H., Barzilai, A., Dror, O., Haspel, N., Nussinov, R., Wolfson, H.J., 2003. Takinggeometry to its edge: fast unbound rigid (and hinge-bent) docking. Proteins 52,107–112.

Schneidman-Duhovny, D., Inbar, Y., Nussinov, R., Wolfson, H.J., 2005. PatchDockand SymmDock: servers for rigid and symmetric docking. Nucleic Acids Res.33, W363–W367.

Sippl, M.J., 1993. Recognition of errors in three-dimensional structures of proteins.Proteins 17, 355–362.

Tensen, C.P., Flier, J., Van Der Raaij-Helmer, E.M., Sampat-Sardjoepersad, S., Van DerSchors, R.C., Leurs, R., Scheper, R.J., Boorsma, D.M., Willemze, R., 1999. Human IP-9: a keratinocyte-derived high affinity CXC-chemokine ligand for the IP-10/Migreceptor (CXCR3). J. Invest. Dermatol. 112, 716–722.

Thompson, J.D., Higgins, D.G., Gibson, T., 1994. CLUSTAL W: improving the sensi-tivity of progressive multiple sequence alignment through sequence weighting,position-specific gap penalties and weight matrix choice. Nucleic Acids Res. 22,4673–4680.

Zeremski, M., Petrovic, L.M., Talal, A.H., 2007. The role of chemokines as inflamma-tory mediators in chronic hepatitis C virus infection. J. Viral Hepat. 14, 675–687.

Zeremski, M., Petrovic, L.M., Chiriboga, L., Brown, Q.B., Yee, H.T., Kinklabwala, M.,Jacobson, I.M., Dimova, R., Markatou, M., Talal, A.H., 2008. Intrahepatic levels ofCXCR3-associated chemokines correlate with liver inflammation and fibrosis inchronic hepatitis C. Hepatology 48, 1440–1450.

Zlotnik, A., Yoshie, O., 2000. Chemokines: a new classification system and their rolein immunity. Immunity 12, 121–127.

Walser, T.C., Rifat, S., Ma, X., Kundu, N., Ward, C., Goloubeva, O., Johnson, M.G.,Medina, J.C., Collins, T.L., Fulton, A.M., 2006. Antagonism of CXCR3 Inhibits

lung metastasis in a murine model of metastatic breast cancer. Cancer Res. 66,7701–7707.

Xanthou, G., Williams, T.J., Pease, J.E., 2003. Molecular characterization of thechemokine receptor CXCR3: evidence for the involvement of distinct extracel-lular domains in a multi-step model of ligand binding and receptor activation.Eur. J. Immunol. 33, 2927–2936.