crossmark THEJOURNALOFBIOLOGICALCHEMISTRY VOL.292,NO.7,pp.2624–2636,February17,2017 ©2017byTheAmericanSocietyforBiochemistryandMolecularBiology,Inc. PublishedintheU.S.A. Inhibition of Mammalian Glycoprotein YKL-40 IDENTIFICATIONOFTHEPHYSIOLOGICALLIGAND*□S Receivedforpublication,October26,2016,andinrevisedform,December22,2016 Published,JBCPapersinPress,January4,2017,DOI10.1074/jbc.M116.764985 AbhishekA.KognoleandX ChristinaM.Payne1 FromtheDepartmentofChemicalandMaterialsEngineering,UniversityofKentucky,Lexington,Kentucky40506 EditedbyGeraldW.Hart YKL-40 is a mammalian glycoprotein associated with pro- easesandamultitudeofcancers(1–4).Manydifferenttypesof gression,severity,andprognosisofchronicinflammatorydis- cellsincludingsynovial,endothelial,epithelial,smoothmuscle, easesandamultitudeofcancers.Despitethiswelldocumented andtumorcellsproduceYKL-40invivo,likelyinresponseto association, identification of the lectin(cid:3)s physiological ligand environmentalcues(5–8).Speculationastobiologicalfunction and,accordingly,biologicalfunctionhasprovenexperimentally ofYKL-40variesfrombothinhibitingandantagonizingcolla- difficult. YKL-40 has been shown to bind chito-oligosaccha- genfibrilformationasaresultofinjuryordisease(9),aswellas rides;however,theproductionofchitinbythehumanbodyhas conferringdrugresistanceandincreasingcellmigrationlead- notyetbeendocumented.Possiblealternativeligandsinclude ing to progression of cancer (3), and protection from chitin- proteoglycans, polysaccharides, and fibers like collagen, all of containingpathogens(10).AlthoughtheassociationofYKL-40 whichmakeuptheextracellularmatrix.ItislikelythatYKL-40is withphysicalmaladiesiswelldocumented,identificationofthe interacting with these alternative polysaccharides or proteins physiologicalligandofthislectin,andthusbiologicalfunction, withinthebody,extendingitsfunctiontocellbiologicalroles remainselusive. such as mediating cellular receptors and cell adhesion and Mammalian YKL-40 is classified as a family 18 glycoside migration.Here,weconsiderthefeasibilityofpolysaccharides, hydrolase based on high homology with this well conserved including cello-oligosaccharides, hyaluronan, heparan sulfate, classofenzymesintheCAZydatabase(10–12).Althoughsim- heparin,andchondroitinsulfate,andcollagen-likepeptidesas ilarinstructuretofamily18glycosidehydrolases,YKL-40lacks physiologicalligandsforYKL-40.Weusemoleculardynamics catalyticactivityasaresultofsubstitutionoftheglutamicacid simulationstoresolvethemolecularlevelrecognitionmecha- andasparticacidmotiftypicalofcatalyticallyactivefamily18 nismsandcalculatethefreeenergyofbindingthehypothesized hydrolases, rendering YKL-40 a lectin, a non-catalytic sugar- ligandstoYKL-40,addressingthermodynamicpreferencerela- binding protein. Structural evidence suggests that YKL-40 tive to chito-oligosaccharides. Our results suggest that chito- exhibitsatleasttwofunctionalbindingregions(10).Thepri- hexaoseandhyaluronanpreferentiallybindtoYKL-40overcol- mary binding cleft has nine binding subsites lined with aro- lagen, and hyaluronan is likely the preferred physiological maticresiduescompatiblewithcarbohydratebinding(Fig.1).A ligand, because the negatively charged hyaluronan shows putativeheparin-bindingsite,locatedwithinasurfaceloop,has enhancedaffinityforYKL-40overneutralchitohexaose.Colla- alsobeensuggested(Fig.1),althoughinvitrobindingaffinity gen binds in two locations at the YKL-40 surface, potentially studieshavebeenunabletoconclusivelydemonstratethis(11). relatedtoaroleinfibrillarformation.Finally,heparinnon-spe- Binding affinity and structural studies reveal that chito- cificallybindsattheYKL-40surface,aspredictedfromstruc- oligosaccharidesarenaturalsubstrates(6,10,11).Inlinewith turalstudies.Overall,YKL-40likelybindsmanynaturalligands family 18 glycoside hydrolases, YKL-40 uniquely binds short in vivo, but its concurrence with physical maladies may be andlongchito-oligomers,indicatingpreferentialsiteselection relatedtoassociatedincreasesinhyaluronan. basedonaffinity(10).Chitohexaosebindinghasalsobeenpur- ported to induce conformational changes in YKL-40 (10), although this has not been observed in all structural studies YKL-40, also known as chitinase 3-like 1, is a mammalian (11). Lectin binding niches are widely believed to be “pre- glycoprotein implicated as a biomarker associated with pro- formed” to the preferred ligand, exhibiting little conforma- gression,severity,andprognosisofchronicinflammatorydis- tionalchangeuponbinding(13,14).Despiteapparentaffinity, chitinisnotanaturalbiopolymerwithinmammalianorbacte- *ThisworkwassupportedbyKentuckyScienceandEngineeringFoundation rialcells,andthepresenceofchitinorchito-oligosaccharidesin Grant KSEF-148-502-13-307). Computational time for this research was mammals is likely related to fungal infection (15). The noted providedbytheExtremeScienceandEngineeringDiscoveryEnvironment (83), which is supported by National Science Foundation Grant ACI- up-regulation of YKL-40 in response to inflammation lends 1053575underAllocationMCB090159.Theauthorsdeclarethattheyhave credencetotheargumentthatYKL-40functionsaspartofthe noconflictsofinterestwiththecontentsofthisarticle. □S This article contains supplemental text, Tables S1–S6, Figs. S1–S8, and innateimmuneresponseinrecognitionofselffromnon-self(6, MoviesS1–S3. 16); however, high expression levels of YKL-40 in carcinoma 1To whom correspondence should be addressed: Dept. of Chemical and tissues suggest function beyond the innate immune response MaterialsEngineering,UniversityofKentucky,177F.PaulAndersonTower, mayalsoexist(17,18).Theextracellularmatrixiscomprisedof Lexington, KY 40506. Tel.: 859-257-2902; Fax: 859-323-1929; E-mail: [email protected]. a mesh of proteoglycans (protein-attached glycosaminogly- This is an Open Access article under the CC BY license. 2624 JOURNALOFBIOLOGICALCHEMISTRY VOLUME292•NUMBER7•FEBRUARY17,2017 IdentifyingthePhysiologicalLigandofYKL-40 therconfoundsthequestionofmechanismwhenconsidering physiological ligands, because YKL-40 is capable of binding bothproteinandcarbohydrates. UnderstandingthemechanismandaffinitybywhichYKL-40 bindsligandsiscrucialtoourunderstandingofitsphysiological function.Thisknowledgewillserveasafoundationforfuture campaigns toward rational development of a potent antago- nistsenablingcellbiologicalstudyandaddressingYKL-40asa therapeutictarget.Toaccomplishthisgoal,wemustdescribe the molecular level mechanisms governing the interaction of YKL-40 with both polysaccharide and collagen-like polypep- tidesandquantitativelyevaluateaffinity.Inthisstudy,weused FIGURE1.SurfacerepresentationofYKL-40showingthebindingcleft classicalmoleculardynamics(MD)2simulationstodifferenti- withaboundhexamerofchitin.Bindingsites(cid:2)2through(cid:3)4arenum- ate modes of ligand recognition and specificity. Using free bered.Sites(cid:3)5,(cid:3)6,and(cid:3)7havealsobeenidentifiedbutarenotshown.The putativeheparin-bindingsiteisshowninblue. energyperturbationwithreplicaexchangemoleculardynamics (FEP/(cid:2)-REMD)andumbrellasamplingMD,wequantitatively cans),polysaccharides,andfibersincludingcollagen.Analter- determinedaffinitiesovercomingtheexperimentaldifficulties nate theory to the pathogenic protection function is that a encounteredthusfar.Asphysiologicalligands,weconsidered closelyrelatedpolysaccharide,insteadofchitin,playstheroleof several options within both the polysaccharide and proteina- thephysiologicalligandinmediatingcellularfunction(11). ceous classes. Below, we provide a brief description of each Despitethestructuralsimilaritybetweenchito-oligosaccha- carbohydrateligandconsidered,aswellasjustificationforcon- ridesandtheproteoglycancarbohydratemonomers,littleevi- siderationofthecollagen-likemodels. denceofpolysaccharidebindingbeyondtheoriginalstructural Polysaccharides—Weselectedpolysaccharidesforthisstudy studiesexists(10,11).Infact,weareawareofonlyoneother basedontheirsimilaritytochito-oligomers,aswellasnatural studyfocusingonthemolecular-levelmechanismofcarbohy- occurrence in mammalian cell walls and/or the extracellular dratebindinginYKL-40(19).Fromabioinformaticsandstruc- matrix. The chito-oligomer from structural studies was turalcomparisonofYKL-40toasimilarchi-lectin,mammary included as a control (10, 11). Chitin is a naturally occurring glandprotein40(20),theauthorsproposeanoligosaccharide biopolymer comprised of repeating N-acetyl-D-glucosamine bindingmechanismthatinvolvestryptophan-mediatedgating (GlcNAc)monomericunitsconnectedby(cid:3)-1,4-glycosidiclink- oftheprimarycarbohydratebindingsite(Fig.1)(19).However, ages(Fig.2).Thecentralringofthemonomer,asix-membered in lieu of a dynamics-based investigation, little can be con- pyranose,iscommontoanumberofcarbohydratesincluding cludedaboutthebindingmechanismofYKL-40ligandsother glucose. Given the chemical similarity, as well as the general thanchito-oligosaccharides,andconformationalchangesrela- presenceofglucoseinmammaliancellsasaformofenergy,a tivetobindingareinaccessible.Fromproteinpurificationtech- hexameric cello-oligomer was also examined as a potential niques, namely heparin-Sepharose chromatography, we also physiologicalligand,despiteitsunlikelypresenceamongmam- knowthatYKL-40reversiblybindsheparin(7,11,21);however, malianglycosaminoglycans. affinitydataforthisinteractiondonotexist.Basedontheinter- Asdescribedabove,YKL-40bindsheparin,andthus,likely action with heparin, it is reasonable to hypothesize heparan alsobindsheparansulfate.Heparansulfate,alesssulfatedform sulfateglycosaminoglycans,existingaspartoftheextracellular ofheparin,isapolysaccharidefoundinabundanceintheextra- matrix construct, are potential physiological ligands. Visual cellularmatrixandatthecellsurface(27).Heparansulfateis inspectionoftheproteinstructuresuggeststhatheparansul- constructed from a repeating disaccharide of (cid:3)-D-glucuronic fatefragmentsmaybeeasiertoaccommodatewithinthecar- acidandN-acetyl-(cid:4)-D-glucosamine(Fig.2).Ofalltheglycos- bohydrate-bindingsitethanheparinitself(11).Itfollowsthat aminoglycans,heparansulfateisthemoststructurallycomplex. otherstructurallysimilarcarbohydratefragmentswouldbind Atleast24differentcombinationsofthedisaccharidemono- withsimilaraffinityinacomparablemechanism. merexist,withdifferencesarisingasaresultofvariationinboth The association of YKL-40 with ailments such as arthritis, isomeranddegreeofsidechainsulfation(28).Additionally,the fibrosis,andjointdiseaseissuggestiveofmolecular-levelinter- heparan sulfate polysaccharide can exhibit both sulfated and actionswithconnectivetissueandthuscollagen(22–26).Moti- unsulfated domains. Physiologically, the unsulfated disaccha- vated by understanding the physiological role of YKL-40 in connectivetissueremodelingandinflammation,Biggetal.(9) ride(cid:3)-D-glucuronicacid(1,4)N-acetyl-(cid:4)-D-glucosamineisthe most prevalent form of heparan sulfate (28). Focusing on the investigatedassociationofYKL-40withcollagentypesI,II,and mostrelevantphysiologicalligands,weexaminedthefullysul- III using affinity chromatography to confirm binding to each fatedformheparinandtheunsulfatedformheparansulfate. type.TheauthorsreportYKL-40specificallybindstoallthree collagentypes.Additionally,theauthorsusedsurfaceplasmon resonancetoconfirmbindingtocollagentypeI.Unfortunately, thereportedaffinityconstantswereinconsistentacrossexper- 2The abbreviations used are: MD, molecular dynamics; PMF, potential of meanforce;FEP,freeenergyperturbation;REMD,replicaexchangemolec- imentsasaresultofaggregation.Nevertheless,theworkclearly ulardynamics;WCA,Weeks-Chandler-Anderson;RMSD,rootmeansquare indicatesthatYKL-40iscapableofbindingcollagen.Thisfur- deviation;RMSF,rootmeansquarefluctuation;PDB,ProteinDataBank. FEBRUARY17,2017•VOLUME292•NUMBER7 JOURNALOFBIOLOGICALCHEMISTRY 2625 IdentifyingthePhysiologicalLigandofYKL-40 Gly-Xaa-Yaa repeating amino acid sequence (34). Generally, the unspecified amino acids, Xaa and Yaa, are proline and hydroxyproline,respectively. Modelcollagenpeptideshavebeenobservedintwodifferent symmetries:theoriginalRichandCrickmodelwith10/3sym- metry (10 units in 3 turns) and the 7/2 symmetry of a more tightly symmetrical triple helix (35–37). On the molecular scale,collagentypewillhaverelativelylittleimpactonbinding toYKL-40.However,symmetrymayhaveanimpactonhydro- genbondinginthebindingsite,andthus,overallaffinity,which will provide unique insight into physiological relevance. To date,modelcollagenpeptidesofatrue10/3symmetryhavenot been reported. Rather, the peptides either have a 7/2 helical pitch or are somewhat “intermediate” in symmetry leading sometobelievethatthe7/2symmetryisrepresentativeofthe true collagen helical structure (38). However, it is not known howuniversallytruethishypothesisisbecausethestructuresof model peptides capture just a small subsection of the larger macromolecularstructure(39). With a broad range of possible collagen architectures, we have selected four representative model collagen peptides FIGURE2.Monomericunitsofthepolysaccharidesconsideredaspoten- whosestructuresarebothavailablefromcrystallographicevi- tialphysiologicalligandsofYKL-40:cellohexaose,chitohexaose,hepa- dence and span the 10/3 and 7/2 symmetries to the greatest ransulfate,heparin,hyaluronan,andchondroitinsulfate.Thechito-oli- gomerisapolymerof(cid:3)-1,4-linkedGlcNAcmonomers.Heparansulfatewas possibleextent.Thefirstcollagenpeptideconsideredisthatof modeledasa(cid:3)-1,4,(cid:4)-1,4-linkedchainofGlcAandGlcNAc.Heparinwasrep- the basic collagen peptide model, PDB code 1CAG (40). The resentedasthe(cid:3)-1,4,(cid:4)-1,4-linkedoligomerofGlcAandGlcNS.Hyaluronan andchondroitinsulfateare(cid:3)-1,3,(cid:3)-1,4-linkedoligomers;theformerconsists reported1.9ÅresolutionstructureexhibitsasingleGlytoAla ofGlcAandGlcNAc,andthelatterconsistsofGlcAandGalNAc.GlcA,(cid:3)-D- substitution and 7/2 symmetry overall. Near the substitution glucuronicacid;IdoA,(cid:4)-D-iduronicacid;GlcNS,N-sulfo-(cid:4)-D-glucosamine;Gal- NAc,N-acetyl-(cid:3)-D-galactosamine. site,thehelixrelaxessomewhatfrom7/2symmetry,although notsomuchastochangeoverallsymmetry.Thesecondcolla- genmodelpeptideweconsiderisavariationofthe1CAGpep- Hyaluronan is a particularly interesting glycosaminoglycan tide, where we reverted the alanine substitution to its native relative to this study, because chito-oligosaccharides are pre- glycine.Minimizationofthisstructurereturnsthehelixtofull cursorstohyaluronansynthesisinvivo(29–31).Thestructural 7/2 symmetry; we refer to this peptide as “native 1CAG” relationshipofthesetwomoleculesissuchthatbindingmech- here. The third model represents a segment from type III anismsmaybesimilaratalternatingbindingsites.Hyaluronan homotrimercollagenwithapproximate10/3symmetryinthe is a polysaccharide of a repeating (cid:3)-D-glucuronic acid and middle part of the helix (PDB code 1BKV) (35). This middle N-acetyl-(cid:3)-D-glucosamine disaccharides connected by alter- part of the 1BKV model peptide, also referred as the T3–785 nating (cid:3)-1,3- and (cid:3)-1,4-glycosidic linkages (Fig. 2). As with peptide, has an imino acid-poor sequence of GITGARGLA. heparansulfate,hyaluronanisalsoaglycosaminoglycancom- Ourfourthmodel,1Q7D,isatriplehelicalcollagen-likepeptide prising the extracellular matrix. At extracellular pH, the car- sequence including a hexapeptide Gly-Phe-Hyp-Gly-Glu-Arg boxyl groups of glucuronic acid are fully ionized, giving the (GFOGER) motif in the middle (41); this motif is not suffi- ligand an overall negative charge under typical physiological ciently long to exhibit 10/3 symmetry, exhibiting, rather, an conditions(32). intermediatedegreeof7/2helicalsymmetry.This1Q7Dmodel Finally,weconsiderchondroitinsulfate,whichisalsoagly- isknowntobindtheintegrin(cid:4)2(cid:3)1-Idomainprotein(42),and cosaminoglycanprevalentinmammals.Theprimarystructural theGFOGERmotifisfoundinthe(cid:4)1chainoftypeIcollagen. unitsofchondroitinsulfatearearepeating(cid:3)-D-glucuronicacid and N-acetyl-(cid:4)-D-galactosamine disaccharides connected by ExperimentalProcedures alternating(cid:3)-1,3-and(cid:3)-1,4-glycosidiclinkages(Fig.2).Aswith MolecularDynamicsSimulations heparansulfate,chondroitinsulfateexistsinvariablysulfated types(33);wehaveselectedtheC4andC6sulfatedvariantof MD simulations were constructed starting from the chito- chondroitinsulfatepolysaccharideasamodel. hexaose-boundYKL-40structure(PDBcode1HJW)deposited Collagen-like Peptide Models—Collagen has also been con- byHoustonetal.(10).Theaposimulationsimplyremovedthe sideredasapotentialphysiologicalligandbasedonthenoted chito-oligomer from the primary binding cleft. The N-linked affinityandparticipationofYKL-40incollagenfibrilformation glycancapturedintheYKL-40structure,atAsn60,wasincluded (9).Collagen,unliketheotherpotentialphysiologicalligands,is in system preparation. Because crystal structures of YKL-40 amacromolecularproteinwithatriplehelixstructure.There boundtootherpolysaccharidesarenotavailable,weusedthe areatleast27distincttypesofhumancollagen,formingavari- structuralsimilarityofpolysaccharidesasthebasisformodel- etyofbiologicalnetworks,allofwhichareconstructedofabasic ingtheremainingpolysaccharidesinthisinvestigation.Inthe 2626 JOURNALOFBIOLOGICALCHEMISTRY VOLUME292•NUMBER7•FEBRUARY17,2017 IdentifyingthePhysiologicalLigandofYKL-40 FIGURE3.Molecularshapecomplementaritydockingcalculationspredictcollagen-likepeptideswillbindtoYKL-40intwopossibleorientations.a, thefrontviewofYKL-40(graysurface)withcollagendockedinsiteA(greenstick).b,thebackviewwherecollagenisdockedinsiteB(cyanstick).c,topviewof YKL-40illustratingrelativepositionsofbindingsites.Theputativeheparin-bindingsubsiteisshowninbluesurfacetoaidinvisualizationofrelativeorientation oftheprotein(cid:4)proteincomplexes.Thisparticularfigureshowstheintegrin-bindingcollagenpeptide,1Q7D(41),inthepredictedbindingsitesalongthesurface ofYKL-40;similardockingwascarriedoutforothercollagenmodels. caseofcellohexaose,hyaluronan,heparansulfate,heparin,and 300K using NAMD (51). Explicit procedural details are pro- chondroitin sulfate, we located the central ring atoms of the videdinthesupplementalmaterials. ligand backbone in the same location as that of the original chitohexaose.Appropriatepyranosesidechainsandglycosidic FreeEnergyCalculations linkages(Fig.2)wereaddedusingCHARMMinternalcoordi- FEP/(cid:2)-REMD—FreeenergyperturbationwithHamiltonian natetablestoconstructtheremainderofthesugarresidue(43). replicaexchangemoleculardynamics(FEP/(cid:2)-REMD)wasused AllpolysaccharidesweredescribedusingtheCHARMM36car- tocalculatetheabsolutefreeenergyofbindingthepolysac- bohydrateforcefield(44–46).Themissingforce-fieldparam- charideligandstoYKL-40(52,53).ThisprotocolusesHam- etersforN-sulfatedglucosamine(supplementalFig.S1)inhep- iltonianreplicaexchangeasameansofimprovingtheBoltz- arin were developed using the force-field Toolkit plugin for mannsamplingoffreeenergyperturbationcalculations.The VMD (47, 48). Details of this parameterization and output parallel/parallel replica exchange MD algorithm in NAMD parameters(supplementalTableS2)arereportedinthesupple- was implemented as recently described (51, 54). The free mentalmaterials. energy calculations performed using this approach were Construction of the collagen-bound YKL-40 models re- accomplished through two separate sets of free energy cal- quired docking calculations to appropriately position the culations following the thermodynamic cycle illustrated in ligand.Thecollagenpeptidesaresignificantlylargerthananyof supplemental Fig. S2. To obtain each binding free energy, thecarbohydrateligands;thus,itisunlikelythatacollagenmol- (cid:4)G,theboundcarbohydrateligandwasfirstdecoupledfrom ecule occupies the primary YKL-40 binding site in the same thesolvatedprotein(cid:4)carbohydratecomplextodetermine(cid:4)G . 1 manneraschito-oligomer.Standardaffinity-baseddockingcal- Thesecondcalculationentaileddecouplingthesolvatedoligo- culations, such as the ones performed in AutoDock, are not saccharidefromsolutionintovacuumtoobtain(cid:4)G .Thedif- 2 feasiblefordeterminationofaninitialcollagen-bindingdomain ferencebetweenthetwovalues,(cid:4)G (cid:3)(cid:4)G ,givestheabsolute 2 1 given the size and flexibility of the triple helix structures. freeenergyofbindingthegivenligandtoYKL-40. Rather, the collagen peptides were docked on the basis of Ineachfreeenergycalculation,fiveseparatetermscontrib- molecularshapecomplementarityusingtheonlinewebserver utetothepotentialenergyofthesystem:thenon-interacting PatchDockBetaversion1.3(49,50).Inthecaseofeachofthe ligandpotentialenergy,repulsiveanddispersivecontributions four collagen-like model peptides, PatchDock predicted two totheLennard-Jonespotential,electrostaticcontributions,and potentialoccupanciesalongthesurfaceofYKL-40,siteAand the restraining potential. In each calculation, the ligand was siteB.BindingsiteAcorrespondstotheprimarycarbohydrate- decoupled from the system by thermodynamic coupling binding domain of YKL-40; however, the collagen ligand was parameters controlling the non-bonded interaction of the notasdeeplyentrenchedinthecleftaschitohexaose.Binding ligand with the environment. The parameters decoupled the siteBislocatedontheoppositesideofYKL-40fromthepri- ligandinafour-stageprocess,whereinthecouplingparameters mary binding cleft. Thus, for each collagen-like peptide, two defined replicas that were exchanged along the length of the MDsimulationswereconstructedrepresentingthetwopoten- alchemical pathway. This decoupling has been described in tialbindingsites.Fig.3illustratestheresultsofthedockingwith detailpreviously(54)andisalsoreportedinthesupplemental predictedcollagenbindingsitesAandBforthe1Q7Dcollagen- materials.Atotalof128FEPreplicaswereused,andaconven- likemodelpeptide. tional Metropolis Monte Carlo exchange criterion governed Theconstructedprotein-ligandsystemswereminimizedin the swaps throughout the replica exchange process (53). The vacuumandsubsequentlysolvatedwithwaterandsodiumions. free energy of binding was determined from 20 consecutive, Using CHARMM (43), the solvated systems were extensively 0.1-ns simulations of each corresponding system, where the minimizedandheatedto300Kfor20ps,whichwasfollowedby first1nsofdatawerediscardedasequilibration.Onestandard MD simulation for 100 ps in the NPT ensemble. The coordi- deviationofthelast1nsofsimulationdatawereusedtoobtain nates following density equilibration were used as a starting anestimateoferror.Additionaldetailsaboutthesecalculations point for 250 ns of MD simulation in the NVT ensemble at aredescribedinthesupplementalmaterials. FEBRUARY17,2017•VOLUME292•NUMBER7 JOURNALOFBIOLOGICALCHEMISTRY 2627 IdentifyingthePhysiologicalLigandofYKL-40 FIGURE4.RelaxationofthepolysaccharideligandsintheprimarybindingcleftofYKL-40.Eachligandisshownaftera100-psequilibrationinathickstick representation.Forcomparison,thechito-oligomer,initsequilibratedconformation,isshowninthingreenlinesbehindeacholigosaccharide.TheYKL-40 proteinhasbeenalignedsuchthateacholigosaccharideisorientedinthesamemanner;althoughYKL-40isnotshownforvisualclarity.Heparansulfate, heparin,andchondroitinsulfaterelaxsignificantlyfromtheinitialdistortedconformation. UmbrellaSampling—ConvergencechallengesmakeFEP/(cid:2)- onlythreebindinastablefashionintheprimarycarbohydrate REMDinappropriatefordeterminingthebindingfreeenergy binding site of YKL-40. The three potential polysaccharide ofthemuchlargerandmoreflexiblecollagen-likemodelpep- physiological ligands at this site include chitohexaose, cello- tides.Thus,umbrellasamplingwasusedtodeterminethework hexaose, and hyaluronan. In the section that follows, we will requiredtodetachthecollagenligandsfromtheshallowclefts describethedynamicsofchitohexaose,cellohexaose,andhya- of YKL-40. Over the entire reaction coordinate, this value luronan binding to YKL-40. The remaining three ligands— equatestobindingaffinity,enablingrelativecomparisonofcol- heparin, heparan sulfate, and chondroitin sulfate—were dis- lagenpeptideaffinity.TheMDumbrellasamplingsimulations lodgedfromthebindingsiteoverthecourseofMDsimulations. usedanativecontacts-basedreactioncoordinateanalogousto The (cid:4)-1,4-glycosidic linkages in heparin and heparan sulfate, that defined by Sheinerman and Brooks (55) and as imple- insteadof(cid:3)-1,4,modifiestherelativeorientationofdisaccha- mented in recent cellulose decrystallization studies (56, 57). ride monomers from that of the chito-oligosaccharide. The Here,anativecontactwasdefinedasaYKL-40proteinresidue NMR solution structure of heparin (PDB code 1HPN) shows within12Åofacollagenpeptideresidue;distancewasdefined thattherelaxedconformationissemi-helical(59),whichcan- by center of geometry of a given residue. The cutoff distance notbefeasiblyaccommodatedintheconserved,narrowcarbo- wasselectedtobelargerthanthenon-bondedcutoffdistance, hydrate binding site of YKL-40. Heparan sulfate suffers from ensuringthatthecollagenligandwasnolongerinteractingwith similarstericconstraintsposedbytherelaxationdrivingforce. YKL-40.Additionally,thewaterboxesofthecollagen-YKL-40 The bulky sulfated side chains of heparin introduce further systemsweremadebiggertoaccommodatetherequiredsepa- steric hindrance and, in the case of heparin and chondroitin rationdistance. sulfate, unfavorably strong electrostatic interactions resulting Thechangeinfreeenergywasdeterminedasafunctionof fromnegativelychargedmoietiesinconvenientlylocatedalong thereactioncoordinate,(cid:5),formulatedastheweightedsumof thecleft(i.e.withoutaco-located,oppositelychargedprotein thestatesofthenativecontacts.Theinitialcoordinatesofthe residue)ejecttheligandsfromthecleft. bound systems were selected from 250-ns equilibrated snap- Inthecasesofheparin,heparansulfate,andchondroitinsul- shots.Theinitialnumberofnativecontactsandtheirweights fate,theligandsquickly“relax”fromtheinitialwideV-shaped werecalculatedfromthesesnapshots.Aninitialreactioncoor- conformationastheyareexpelledfromthecleftbycharge-and dinateof0(normalized)correspondstothisinitialcondition, steric-based effects. Relaxation of the sugar from the initial and a final reaction coordinate of 1 corresponds to all of the bindingposeissufficienttoinitiatelossofcriticalnon-bonded nativecontactsbeingoutsidethe12-Åcutoff(i.e.theligandis interactionsalongwithasubsequentreductioninaffinity(Fig. decoupledandfreelysamplingthebulk).Thereactioncoordi- 4).Within25ns,heparin,heparansulfate,andchondroitinsul- natewasdividedinto20windowsevenlyspacedalongthereac- fatewereexpelledfromthecleftintobulksolution.Eachofthe tioncoordinate,andeachwindowwassampledfor5ns,where threeligandscapableofbindingwiththeprimarybindingcleft thereactioncoordinatewasmaintainedatthespecifiedvalue maintainedthe(cid:3)1boatconformationovertheentiresimula- usingaharmonicbiasingforcewiththeforceconstantof41,840 kJ/mol. The potentials of mean force profiles were calculated tion.Chitohexaoseandcellohexaoseremainedinthebinding usingtheweightedhistogramanalysismethod(WHAMsoft- cleftovertheentire250-nsMDsimulationwhilemaintaining ware),anderroranalysiswasperformedusingbootstrapping. theinitialwideVshape.HyaluronandevelopedasharpVshape Adetailedlistingofallthesimulationsandfreeenergycalcula- within a few nanoseconds and maintained this conformation tions performed as part of the objectives of this study is pro- within the binding cleft for the remainder of the simulation videdinsupplementalTableS1. (supplementalFig.S6);thisisprimarilyduetovariationingly- cosidic linkage, where hyaluronan exhibits a (cid:3)-1,3-linkage ResultsandDiscussion withinthemonomerinsteadofthe(cid:3)-1,4-linkageofcello-and Protein-PolysaccharideBindinginYKL-40—MDsimulation chitohexaose. Also, comparison of the equilibrated chito- suggeststhatofthesixpolysaccharideoligomersinvestigated, hexaose- and hyaluronan-bound structures disabuses one of 2628 JOURNALOFBIOLOGICALCHEMISTRY VOLUME292•NUMBER7•FEBRUARY17,2017 IdentifyingthePhysiologicalLigandofYKL-40 FIGURE5.SnapshotsfromfourindependentMDsimulationsofheparin(whitestick)bindingtoaputativeheparin-bindingsite(bluesurface)ofYKL-40 (graysurface).TheprimaryoligosaccharidebindingsiteofYKL-40ismarkedbyanaromaticresidueshowninsalmonsurfacerepresentation.Transparent spheresillustratetheinitialsimulationpositionsofheparin.Intwocases,aandb,theheparinligandwasinitiallyboundintheprimaryYKL-40bindingsite.In bothcases,theligandwasexpelledfromtheprimarybindingsiteintosolutionandlocatedtheheparin-bindingsitethroughelectrostaticinteractions.Two additionalsimulations,candd,wereinitializedwiththeheparinligandfreeinsolution.Again,theligandsidentifiedtheYKL-40heparin-bindingdomain throughcharge-basedinteractions.Theliganddidnotspecificallybindinaparticularconformation.Rather,theliganddynamicallyinteractedwiththe heparin-bindingdomain. thenotionthatsimilarbindingmechanismsexistatalternate dentMDsimulationsoftheYKL-40/heparinsystem:onewitha binding sites, because only the (cid:3)1 site pyranose appears to newrandomnumberseed,althoughinthesameconfiguration, maintainsimilarsidechainorientation. andtwoadditionalsimulationswiththeligandrandomlyplaced Thenativedistortedconformationischaracteristicofglyco- insolution(supplementalMovieS2).Ineachcase,theheparin sidehydrolasepyranosebindingbehaviorinthe(cid:3)1site(Fig.4) oligomers were capable of finding and binding to a group (60). In solution, polysaccharide pyranose moieties adopt the of charged residues at the surface of YKL-40 (Fig. 5); these energeticallyfavorablechairconformation(61);however,when were the basic residues of a putative heparin-binding site, boundtoanenzyme,theactivesitesofcatalyticallyactivegly- GRRDKQH, at positions 143–149. Interestingly, this domain cosidehydrolasesdistortthepyranoseringinthe(cid:3)1binding followsaconsensusproteinsequence,XBBXBX(whereBisa subsiteintoalessenergeticallyfavorableconformation,suchas basicresidueandXisanynon-basicaminoacid),thatisnoted aboatorskewconformation(62–65),primingthesubstratefor for its ability to recognize polyanions like heparin (67). In all hydrolytic cleavage. Interestingly, the chitohexaose ligand fourcases,heparinrecognizedthebindingsitewithin25nsof bound in the primary binding site of YKL-40 exhibits a boat MDsimulation(Fig.5),occasionallyvisitingothermoderately conformation despite not being catalytically active (10). This basicsurfacelocationsbeforelocalizingaroundtheGRRDKQH suggeststhatthesugardistortioninthe(cid:3)1bindingsitecon- motif. The strong electrostatic interaction arose from the tributestoligandbindingaswellascatalysis,becausethereisno dynamicformationofsaltbridgesbetweeneitherthesulfateor evolutionary requirement to overcome an activation energy thecarboxylgroupsoftheheparinoligosaccharideandtheside barrierincatalyticallyinactivelectins.Arecentstudyofaho- mologous chitinase suggested that (cid:3)1 pyranose relaxation chainsofthebasicaminoacids(supplementalTableS3).Lys155 andLys193alsoexhibitlarge,favorableinteractionswithhepa- reducesbindingaffinityandaffordstheligandmoreflexibility rin, interacting with the opposite end of the polysaccharide andentropicfreedom(66),whichisconsistentwithourfind- ligand as it binds at the surface. Coupled with experimental ingsfromthe250-nsMDhere. PutativeHeparin-bindingSite—Despitethefactthatthehep- observationofheparinaffinity,ourMDsimulationssuggesta arinoligomercouldnotbeaccommodatedbytheYKL-40bind- non-specific, surface-mediated binding interaction between ingcleft,MDsimulationsdosuggestthattheoligomerinteracts YKL-40andtheextensivelysulfatedheparinoligomer(10,11). with the surface of YKL-40 at a putative heparin-binding site Althoughtheunsulfatedvariant,heparansulfate,didnotvisit (Fig. 1). After ejection from the primary binding site, the oli- the heparin-binding site, chondroitin sulfate also attached to gomerspontaneouslybindstotheYKL-40heparin-bindingsite theputativeheparin-bindingsiteinasimilarfashiontoheparin. (supplemental Movie S1). To address the significance of this Giventhechemicalsimilarityoftheseglycosaminoglycans,i.e. unanticipated event, we performed three additional indepen- highly sulfated and negatively charged, we anticipate the FEBRUARY17,2017•VOLUME292•NUMBER7 JOURNALOFBIOLOGICALCHEMISTRY 2629 IdentifyingthePhysiologicalLigandofYKL-40 TABLE1 EnergeticcomponentsofthefreeenergyofligandbindingtoYKL-40 System (cid:2)G (cid:2)G (cid:2)G (cid:2)G (cid:2)G (cid:2)G repu disp elec rstr Tot b kJmol(cid:3)1 kJmol(cid:3)1 kJmol(cid:3)1 kJmol(cid:3)1 kJmol(cid:3)1 kJmol(cid:3)1 YKL-40(cid:2)Cellohexaose 379.5(cid:5)3.8 (cid:3)382.3(cid:5)2.1 (cid:3)306.6(cid:5)2.2 (cid:3)1.1 (cid:3)310.5(cid:5)3.6 (cid:3)12.6(cid:5)3.7 Cellohexaose 316.2(cid:5)1.8 (cid:3)287.4(cid:5)1.3 (cid:3)326.7(cid:5)1.6 0 (cid:3)297.9(cid:5)2.8 YKL-40(cid:2)Chitohexaose 535.9(cid:5)11.7 (cid:3)538.9(cid:5)4.2 (cid:3)407.7(cid:5)7.5 (cid:3)5.5 (cid:3)416.2(cid:5)11.2 (cid:3)63.3(cid:5)12.6 Chitohexaosea 329.6(cid:5)4.5 (cid:3)306.2(cid:5)3.0 (cid:3)376.3(cid:5)3.4 0 (cid:3)352.9(cid:5)5.9 YKL-40(cid:2)Hyaluronan 438.1(cid:5)2.8 (cid:3)439.5(cid:5)1.5 (cid:3)1395.9(cid:5)4.2 (cid:3)1.3 (cid:3)1398.6(cid:5)4.3 (cid:3)106.8(cid:5)4.6 Hyaluronan 333.1(cid:5)1.4 (cid:3)306.0(cid:5)1.3 (cid:3)1318.9(cid:5)1.5 0 (cid:3)1291.8(cid:5)2.5 aHamreetal.(68). XBBXBXmotifmayalsoroutinelyappearinchondroitin-bind- charideligandcapableofbindingintheYKL-40bindingcleft ingproteins. willlikelyexhibitasimilarlyfavorableWCAbindingfreeenergy Polysaccharide Ligand Binding Affinity—Each of the three component. In the following section, we expand upon the polysaccharidesmaintainingcontactwiththeprimarybinding molecularlevelinteractionsthatcontributetopolysaccharide siteofYKL-40,cellohexaose(orlikelyanyglucosederivative), bindingaffinityinYKL-40. chitohexaose, and hyaluronan, are feasible ligands. However, Based on these results, it is unlikely that a cello-oligomer freeenergycalculationssuggestthathyaluronanmaypreferen- wouldbindinthecleftofYKL-40overachito-oligomer,and tially bind with YKL-40 if chitin is not present in the human thus, although there is potential for YKL-40 to bind a cello- bodyasaresultoffungalinfection.Theabsolutefreeenergiesof oligomerorglucose,itwouldnotbeinhibitory.Hyaluronan,on bindingcellohexaose,chitohexaose,andhyaluronantoYKL-40 theotherhand,likelycompeteswithchito-oligomersinbind- were (cid:3)12.6 (cid:5) 3.7, (cid:3)63.3 (cid:5) 12.6, and (cid:3)106.8 (cid:5) 4.6 kJ/mol, ing,whichisdueinlargeparttotheelectrostaticfavorabilityof respectively. Repulsive, dispersive, and electrostatic compo- thechargedsidechainsofhyaluronanintheYKL-40binding nentsofthefreeenergychangesaretabulatedinTable1.Con- cleft.Clinicaldatasupporthyaluronanasabiomarkerforcan- vergence assessment of the free energy calculations has been cer prognosis and inflammation (32, 70), the same events in provided in supplemental Fig S3. We recently calculated the which YKL-40 appears at elevated serum levels (5). To our freeenergyofsolvationforchitohexaoseaspartofastudyon knowledge,therearenostudiesevaluatingtheco-existenceof family18chitinases(68);thisvaluehasbeenusedinourcalcu- YKL-40 and hyaluronan. The cell receptor protein CD44 has lationofchitohexaosebindingaffinitytoYKL-40forcomputa- been implicated in hyaluronan binding interactions and is tionalefficiency.Themethodsusedtocalculatesolvationfree also involved in confounding scenarios, both aggravating and energyofchitohexaosewereidenticaltothosedescribedhere. improvinginflammation(71).SequencealignmentofYKL-40 Furthermore, the binding free energy of chitohexaose to with the hyaluronan-binding domain of human CD44 (72), YKL-40isingoodagreementwiththatofhomologousfamily18 usingBLASTP2.3.0(73),showsnohomology,furthersuggest- chitinases,despitemutationofthecatalyticmotifinthelectin. ingthatthisYKL-40-hyaluronanbindingisdifferentfrompre- Chitohexaoseandcellohexaosearebothneutralligandsbut viouslyknownhyaluronan-bindingproteins(74). display a significant difference in binding affinity to YKL-40. PolysaccharideBindingDynamics—YKL-40ishighlyhomo- Electrostaticinteractionsappeartobeoneofthemoresignifi- logous with carbohydrate-active enzymes found in glycoside cantcontributorstotheenhancedaffinityofchitohexaoseover hydrolasefamily18(12,75).Despitelackingcatalyticability,the cellohexaose(Table1).Forcellohexaose,thechangeintheelec- primary polysaccharide binding site of YKL-40 exhibits trostatic component of binding free energy was unfavorable remarkablesimilaritytothesefamily18chitinases.Assuch,one (20.1 (cid:5) 2.7 kJ/mol), whereas the same component for chito- may reasonably expect that ligand binding within this family hexaose was energetically favorable ((cid:3)31.4 (cid:5) 8.2 kJ/mol). In willdemonstratesimilartrends,regardlessofevolutionaryori- thecaseofhyaluronan,electrostaticinteractionsplayaneven gin. Indeed, we observe that chitohexaose, cellohexaose, and greater role in enhancing affinity of the ligand for YKL-40 hyaluronan binding in the primary binding site of YKL-40 ((cid:3)77.0 (cid:5) 4.5 kJ/mol). This increasing electrostatic contribu- follow a general pattern common to carbohydrate-active tionisreflectiveofincreasingnumberofelectronegativeatoms enzymes. Namely, that ligand binding interactions are medi- inthesidechainsofcarbohydratesaswegofromcellopentaose ated by carbohydrate-(cid:6)stacking interactions with aromatic tochitohexaosetohyaluronan.Weobservenosignificantdif- residues, and hydrogen bonding interactions are critical to ferencesincellohexaose,chitohexaose,orhyaluronanbinding overallligandaffinityandstability.Weinvestigatethesetrends to YKL-40 arising from Weeks-Chandler-Anderson (WCA) quantitativelythroughanalysisoftheMDsimulationtrajecto- dispersionandrepulsion(Table1).Thisislargelyafunctionof ries,includingrootmeansquaredeviation(RMSD)ofthepro- themolecularsimilarityofthepyranoseringscomprisingthe tein,rootmeansquarefluctuation(RMSF)ofboththeprotein monomericunitsofthreeoligosaccharides(Fig.2).Thepyra- and the ligands over the course of the simulation, hydrogen noseringsofcarbohydratesboundintheactivesitesofglyco- bondinganalysis,degreeofsolvationoftheligand,andinterac- sidehydrolases,andbyextensionthebindingcleftsoflectins, tionenergyoftheligandwiththeprotein. form carbohydrate-(cid:6)stacking interactions with surrounding Cellohexaose,chitohexaose,andhyaluronanbindinginthe aromaticresiduesalongtheclefts(69).InYKL-40,thesestack- primary YKL-40 binding site did not adversely affect protein inginteractionsareformedinthe(cid:3)3and(cid:3)1bindingsiteswith dynamics.Ineachcase,bindingthepolysaccharideliganddid residuesTrp31andTrp352,respectively.Naturally,anypolysac- notsignificantlydisturbtheproteinbackbone(i.e.proteinfold), 2630 JOURNALOFBIOLOGICALCHEMISTRY VOLUME292•NUMBER7•FEBRUARY17,2017 IdentifyingthePhysiologicalLigandofYKL-40 luronanisasstable,ifnotmoreso,aschitohexaoseinthepri- marybindingsite(Fig.6a);however,cellohexaoseappearstobe morestablerelativetothetwootherligandsattheendsofthe bindingcleft.Thislatterfindingisafunctionofthelengthofthe sidechainsattachedtothepyranoseringsofeachoftheligands. Ofthethreecarbohydrates,thecello-oligomerhastheshortest sidechains,andthus,theligandfluctuateslessbecauseitdoes not need to rearrange as significantly to induce binding. As shownabove,thisdoesnotnecessarilycorrespondtothemost thermodynamicallypreferentialligand,andlowerRMSFcould also be interpreted hypothetically as loss of translational and conformational freedom, resulting in unfavorable entropic contribution. The hydrogen-bonding partners of chitohexaose, cello- hexaose,andhyaluronanarequitedifferent,largelyasaresult of the conformational change of hyaluronan (supplemental TableS4).Definingahydrogenbondasapolaratomwithin3.4 Å and 60° of a donor, we identified the formation of donor- acceptorpairsandpercentoccupancyofthesehydrogenbonds between the protein and each carbohydrate moiety over the courseofthe250-nsMDsimulations(supplementalTableS4). Asdescribedabove,hyaluronanformedasharpVshapeinthe polysaccharidebindingcleft,whichminimizedsterichindrance and, in turn, modified accessible hydrogen bonding partners relativetochito-andcellohexaose.Hydrogenbondsatthe(cid:2)1 site, between glucuronic acid and Asp207, Arg263, and Tyr141, stabilized the hyaluronan conformation (supplemental Table S4).Inthe(cid:3)1subsite,chitohexaoseprimarilyhydrogenbonds withthesidechainofTyr206andthemainchainofTrp99.Inthe casesofbothcellohexaoseandhyaluronan,theinteractionwith Tyr206wasabolishedandinsteadsupplementedbyTrp99alone. Inthe(cid:3)2subsite,theoxygenofthechitohexaoseacetylformsa long-livedhydrogenbondwiththeindolenitrogenofthebur- iedTrp352;neitherhyaluronannorcellohexaoseinteractwith the (cid:3)2 site through this tryptophan. Rather, Trp31, which stackswiththepyranoseinthe(cid:3)3subsite,actsasahydrogen donortothe(cid:3)2subsiteglucuronicacidsidechainofhyaluro- nan.Inthecaseofcellohexaose,themainchainofasolvent- FIGURE6.a,RMSFofthepolysaccharideligandsonaper-binding-subsite exposed asparagine, Asn100, almost exclusively mediates basis.Theerrorbarswerecalculatedusingblockaveragingover2.5ns.b, hydrogenbondinginthe(cid:3)2site.The(cid:2)2,(cid:3)3,and(cid:3)4binding averagenumberofwatermoleculeswithin3.5Åofeachligandmonomer. subsites exhibit less frequent hydrogen bonding between the Theerrorbarsrepresentonestandarddeviation. ligandandtheprotein,andthereislittleconsistencyinbonding andtheligandremainedrelativelyunperturbedoverthecourse partnersacrossligands.Certainly,thesevariationswillmanifest ofthesimulation.TherelativelyconsistentRMSDofthepro- in enthalpic contributions to ligand binding, because even a tein backbones suggests that the simulations reached a local singlehydrogenbondmayaccountfor4–29kJ/molofbinding equilibrium(supplementalFig.S4a).AswiththeRMSDcalcu- freeenergyinbiologicalsystems(76,77);suchislikelythecase lation, the RMSF of the protein backbone suggests that the for cellohexaose and chitohexaose binding to YKL-40, where bindingofpolysaccharidesdoeslittletodisturbtheoverallpro- the latter exhibits both greater hydrogen bonding capability tein conformation (supplemental Fig. S4b). Additionally, we andamorefavorablebindingfreeenergy. didnotobservesignificantconformationalchangesinthecar- Keyaromaticresidues,Trp31,Trp99,andTrp352,playasig- bohydrate-bindingsiteofYKL-40beforeorafterligandbinding nificant role in binding all three oligosaccharides. Notably, (supplementalFig.S5).Moredetailsofproteindynamicsand these tryptophans are conserved in other lectins, including conformationalchangeshavebeenprovidedinthesupplemen- mammaryglandprotein(MGP-40)andmammalianlectinYm1 taltext. (20, 78). According to previous structural studies, these aro- TheRMSFoftheligand,averagedover250nsasafunctionof matic residues form hydrophobic stacking interactions with binding site, provides a measure of relative ligand stability. pyranose moieties at the (cid:3)3, (cid:2)1, and (cid:3)1 binding subsites, Error was estimated by block averaging over 2.5-ns blocks. respectively (11). As mentioned above, this carbohydrate-(cid:6) Ligandstabilityoverthecourseofthesimulationsuggestshya- stackingwasobservedacrossthethreepolysaccharideligands FEBRUARY17,2017•VOLUME292•NUMBER7 JOURNALOFBIOLOGICALCHEMISTRY 2631 IdentifyingthePhysiologicalLigandofYKL-40 FIGURE7.Nativecontactanalysisofeachcollagen-likepeptidemodelbindingtoYKL-40atsiteA.Thecolorscalerepresentsthenormalizedfrequency (i.e.fractionalpercentageofframesinwhichthecontactwasformed)oftherespectiveYKL-40residueasanativecontact.Anativecontactwasdefinedasany timeacollagenresiduewaswithin12ÅofaYKL-40residue,wheredistancewasdefinedbycenterofgeometryofagivenresidue.Onlyframesfromthelast100 nsofsimulation,followingaperiodofequilibration,wereconsideredinthisanalysis. asaresultofthechemicallysimilarcarbohydrate“backbone”of passesasmanypotentialbindingmodesasfeasibletodescribe pyranose rings. However, at the (cid:2)1 site of hyaluronan, the protein-protein binding dynamics and relative affinity of stacking interaction with Trp99 was not maintained. Instead, YKL-40forcollagen. prominenthydrogenbondingforcesthe(cid:2)1pyranoseringinan LigandBindingDynamicsandComparisonofModelCol- orientation that is perpendicular to aromatic Trp99 (supple- lagen-like Peptides—Dynamics of the collagen-like peptide mentalFig.S6).Nevertheless,thesimilarityinWCAcontribu- ligandsvarieswithbothbindingsiteandthepitchofthetriple tionstobindingfreeenergyforallthreepolysaccharidessug- helix.Rootmeansquaredeviationillustratestherelativestabil- geststhatthis(cid:2)1sitestackinginteractionweaklycontributes ityofeachcollagenpeptideineachofthetwobindingsites,A totheoverallbindingfreeenergy. andB(supplementalFig.S7).Althoughthemoleculardocking ThedegreetowhichthebindingcleftofYKL-40wasacces- resultsinveryclosecontactbetweencollagenandYKL-40(Fig. sibletowatermoleculesdidnotchangesignificantlywiththe 4), such that collagen appears to be almost buried in the pri- bound polysaccharide. The degree of ligand solvation was marycarbohydrate-bindingsiteofYKL-40,minimizationand determinedbycalculatingtheaveragenumberofwatermole- MDsimulationresultsintheslightriseandshiftintheposition culeswithin3.5Åofagivenpyranoseringoverthecourseofthe ofcollagenforeverymodelatbindingsiteA.Eachofthefour simulations(Fig.6b);errorisgivenasonestandarddeviation. ligandsmaintainsassociationwiththebindingsiteAoverthe Chitohexaoseandcellohexaosedisplaysimilardegreesofsol- courseof250ns,althoughwithslightlydifferentprotein-pro- vationacrossthelengthofthecleft.Hyaluronanallowsamod- tein contacts with YKL-40 (Fig. 7). Native 1CAG, 1BKV, and erateincreaseindegreeofsolvationofthecleftbycomparison 1Q7Dattainedrelativestabilityinapositionnotsignificantly tochitohexaose,acrossthe(cid:3)3and(cid:2)1subsites,whereitssharp differentfromtheinitialdockedposition,butthe1CAGpep- Vshapeagaincontributestovariationinbehavior.Giventhe tide (supplemental Movie S3), with disrupted helical content similarity in solvent accessibility within the binding cleft resultingfromtheglycinetoalaninemutation,requiredanadjust- regardless of ligand, it is unlikely that entropic contributions mentinpitchbeforeassociatingwithYKL-40.Thisrelativechange fromsolvationplayaroleintheobserveddifferencesinligand in position is shown in the RMSD of the peptides during first bindingfreeenergy. 50–100nsbeforestabilization(supplementalFig.S7a).Binding Protein-Protein Binding in YKL-40—Based on biochemical site B accommodates helical pitches of 7/2 collagen peptides, characterization,itisclearthatYKL-40functionallyinteracts becausenative1CAGand1Q7DassociatedwithYKL-40with withcollagen.Forexample,Biggetal.(9)uncoveredtheability very little change in orientation relative to the initial docked ofYKL-40tospecificallybindtypesI,II,andIIIcollagenfibers, positions.The1CAGligandwasexpelledfrombindingsiteB,as andIwataetal.(79)recentlydiscoveredthatYKL-40secreted wasthesomewhatimperfect10/3-pitched1BKVpeptide.This by adipose tissue inhibits degradation of type I collagen by suggeststhatYKL-40mayavoidphysiologicalinteractionswith matrix metalloproteinase-1 and further stimulates the rate of certaincollagenfibrildomains,especiallythosehavingimper- type I collagen formation. However, a lack of structural evi- fecthelicalpitches.Theintegrin-bindingcollagen-likepeptide dencehasprecludeddevelopmentofanunderstandingofthe 1Q7Ddemonstratedthegreateststabilityamongcollagenpep- molecularnatureoftheseinteractions.Usingmoleculardock- tidesinbothbindingsites(supplementalFig.S7)andformed ing,MDsimulation,andfreeenergycalculations,wedescribe morenativecontactswithYKL-40thantheotherthreecollagen interactions of four collagen-like peptides with two putative peptides at binding site A (Fig. 7). We anticipate that the protein-binding sites along the surface of YKL-40. The selec- GFOGERmotifplayssubstantialroleinmediatingtheinterac- tionofmodelpeptides,aswellasmultiplebindingsites,encom- tionofthiscollagenpeptidewithYKL-40. 2632 JOURNALOFBIOLOGICALCHEMISTRY VOLUME292•NUMBER7•FEBRUARY17,2017 IdentifyingthePhysiologicalLigandofYKL-40 Examiningthenumberofnativecontactsbetweeneachcol- lagenpeptideandbindingsiteAofYKL-40revealsseveralcom- moninteractionsitesmediatecollagenbindingandhelpsnar- row down key regions of interest (Fig. 7). YKL-40 residues 69–71, 98–108, 205–215, and 230–235 interact with all four collagenpeptidesandlikelycontributetobindingaffinity,aswe willdiscussbelow.TheregionofYKL-40betweenresidues179 and189associateswithnative1CAG,1BKV,and1Q7D,butnot withtheoriginal1CAG,asthispeptidewithrelaxedsymmetry neededtoadjustitspositionfromdockedconformationtosta- bilize the interactions. The 1Q7D model formed the greatest number of interactions with YKL-40 residues relative to the otherthreemodels.Similarnativecontactanalysisforbinding siteBshowsthatevenN-terminalandC-terminalresiduesof YKL-40areinvolvedincollagenbindingatbindingsiteB(sup- plementalFig.S8).Itshowsthat,unlikebindingsiteA,thereis littledifferenceinthenumberofinteractionsofmodel1Q7D and native 1CAG collagen peptide with the binding site B of YKL-40. To better understand the interactions collagen makes with YKL-40, identified through the native contact analysis, we calculatedelectrostaticandvanderWaalsinteractionener- giesofeachYKL-40residuewitheachcollagenpeptideover the250-nsMDsimulations(supplementalTableS5).Visual inspectionofthesimulationsrevealsaromaticresiduesinthe binding sites, such as Trp212 and Trp99 in binding site A and Phe49 in binding site B, were involved in aromatic-proline stacking interactions with the collagen triple helices. Such interactions are favorable, occurring because of both hydro- phobiceffectsandinteractionbetweenthe(cid:6)aromaticfaceand thepolarizedC-Hbonds(80).Thisisillustratedinthevander Waalscomponentoftheinteractionenergy,whereatbinding siteA,Trp69,Trp71,Trp99,Trp212,andPhe234showsubstantial favorableinteractionwithcollagenpeptides,althoughthecon- tribution varies with each collagen peptide (supplemental TableS5).Additionally,acidicandbasicresiduesoftheinteg- rin-bindingGFOGERmotiffromcollagen-likepeptide1Q7D FIGURE8.CollagenbindingwithYKL-40.a,saltbridgesformedbetween formionicinteractionswiththecounter-ionicaminoacidsof the1Q7Dcollagenpeptide(greencartoon)andbindingsiteA(graysurface).b, saltbridgeinteractionsofthe1Q7Dcollagenpeptide(cyancartoon)with YKL-40, also known as salt bridges. Specifically, 1Q7D forms bindingsiteB(graysurface).c,bindingfreeenergyobtainedfromumbrella threesaltbridgesatbindingsiteAandonesaltbridgeatbinding sampling MD simulations of the YKL-40-collagen peptide systems, inter- siteB(Fig.8).AtsiteA,Arg105,Asp207,andArg263ofYKL-40 pretedasnegativeofPMFtodecouplethepartners.Thecollagenpeptidesin questionare1Q7D(atbothsites)and1CAG(atsiteAonly).Thefreeenergyis interact with Glu(a11), Arg(c12), and Glu(c11) of 1Q7D, shownasafunctionofnormalizedreactioncoordinate,wherethereaction respectively,wherethea,b,orcintheparenthesescorresponds coordinateisfractionofnativecontacts. to one of the three strands of the collagen model. Notably, Glu(a11),Arg(c12),andGlu(c11)belongtotheGFOGERinteg- staticinteractionenergies;supplementalTableS5),althoughas rin-bindingmotif.AtsiteB,Lys23ofYKL-40formsasaltbridge aresultofhydrogenbondingratherthansaltbridgeformation. with Glu(a11) of 1Q7D. As anticipated, the GFOGER motif WenotethattheinteractionenergiesofYKL-40residueswith playedasubstantialroleintheinteractionofthiscollagenpep- residuesofeachcollagenmodelpeptidearenotconservedasa tide with YKL-40, but its role was different from that of the resultofdifferencesinthecollagensequences,particularlyin integrinbindingmechanism,whichfurtherinvolvescoordina- themiddleregionsconsistingofdifferentimino-triplets. tionofametalion(42).Nevertheless,saltbridgesandhydro- From MD simulation, we observe substantial hydrogen phobiccontactsareveryimportantinbothcases,significantly bondingbetweenthecollagenpeptidesandYKL-40acrossthe contributing to the electrostatic component of the binding lengthofeachbindingsite,whichcontributestooverallstabil- affinity of this collagen peptide relative to collagen peptides ityandbindingaffinity.Thehydrogenbondinganalysisforthe lackingacidicorbasicaminoacids(e.g.native1CAG)(supple- collagen-YKL-40systemswasperformedasdescribedabovefor mental Table S5). The hydroxyl oxygens of hydroxyprolines the polysaccharide ligands; pairs exhibiting greater than 10% from 1CAG and native 1CAG appear to be involved in ionic occupancy over the simulation are reported individually in interactions with acidic YKL-40 residues (favorable electro- supplemental Table S6. The YKL-40 residues responsible for FEBRUARY17,2017•VOLUME292•NUMBER7 JOURNALOFBIOLOGICALCHEMISTRY 2633