Enzymatic hydrolysis studies of arabinogalactan-protein structure from Acacia gum: the self-similarity hypothesis of assembly from a common building block

HAL Id: hal-01837409

https://hal.archives-ouvertes.fr/hal-01837409

Submitted on 27 May 2020

HAL is a multi-disciplinary open access

archive for the deposit and dissemination of

sci-entific research documents, whether they are

pub-lished or not. The documents may come from

teaching and research institutions in France or

abroad, or from public or private research centers.

L’archive ouverte pluridisciplinaire HAL, est

destinée au dépôt et à la diffusion de documents

scientifiques de niveau recherche, publiés ou non,

émanant des établissements d’enseignement et de

recherche français ou étrangers, des laboratoires

publics ou privés.

Enzymatic hydrolysis studies of arabinogalactan-protein

structure from Acacia gum: the self-similarity

hypothesis of assembly from a common building block

Denis Renard, Laurence Lavenant-Gourgon, Alain Lapp, Michael Nigen,

Christian Sanchez

To cite this version:

Denis Renard, Laurence Lavenant-Gourgon, Alain Lapp, Michael Nigen, Christian Sanchez.

Enzy-matic hydrolysis studies of arabinogalactan-protein structure from Acacia gum: the self-similarity

hypothesis of assembly from a common building block. Carbohydrate Polymers, Elsevier, 2014, 112,

pp.648-661. �10.1016/j.carbpol.2014.06.041�. �hal-01837409�

ContentslistsavailableatScienceDirect

Carbohydrate

Polymers

jo u r n al h om ep age :w w w . e l s e v i e r . c o m / l o c a t e / c a r b p o l

Enzymatic

hydrolysis

studies

of

arabinogalactan-protein

structure

from

Acacia

gum:

The

self-similarity

hypothesis

of

assembly

from

a

common

building

block

D.

Renard

_,

_L.

_{Lavenant-Gourgeon}

_,

_A.

_Lapp

_,

_M.

_Nigen

_,

_C.

_Sanchez

a_{INRA,UR1268BiopolymèresInteractionsAssemblages,F-44300Nantes,France} b_{LaboratoireLéonBrillouin,CEASaclay,F-91191Gif-sur-Yvettecedex,France}

c_{UMR1208IngénieriedesAgropolymèresetTechnologiesEmergentes,INRA-MontpellierSupAgro-CIRAD-UniversitéMontpellier,2PlacePierreViala,}

F-34060Montpellier,France

a

r

t

i

c

l

e

i

n

f

o

Received25April2014

Receivedinrevisedform16June2014 Accepted18June2014

Availableonline24June2014 Keywords:

HPSEC-MALLS

Small-angleneutronscattering Circulardichroism

Arabinogalactan-protein Acaciagum

Self-ordering

a

b

s

t

r

a

c

t

Arabinogalactan(AG)andarabinogalactan-protein(AGP)fractionsweretreatedenzymaticallyusing sev-eralproteasesinacidic(pH4)andalkaline(pH7)conditionsinordertogodeeperinsightintothestructure andconformationsofthetwomainfractionsofAcaciasenegalgum.EndoproteinaseGlu-C,pepsinand phosphataseacidwerethususedinacidicconditionswhilesubtilisinA,pronase,trypsin,papainand proteinaseKwereusedinalkalineconditionstocleaveproteinmoietiesofthetwofractions.Structures ofAGandAGPwereprobedusingHPSEC-MALLS,smallangleneutronscatteringandfar-UVcircular dichroism.

EnzymesdidnotaffectAGfractionstructurewhateverthepHconditionsused,highlightingthe inac-cessibilityofthepeptidebackboneandtheremarkablestabilityofthisfractioninacidicandalkaline conditions.Thisresultwasinagreementwiththethinoblateellipsoidmodelwepreviouslyidentifiedfor theAGfractionwherethe43amino-acidresiduespeptidesequencewassupposed,basedonspectroscopic methods,tobetotallyburied.

ContrarytoAGfraction,AGPproteincomponentwasthereforecleavedusingenzymesinalkaline conditions,theabsenceofenzymaticefficiencyinacidicconditionsbeingprobablyascribedtolongrange electrostaticrepulsionsoccurringbetweennegativelychargedAGPandenzymesatpH4.Thedecrease ofAGPmolecularweightafterhydrolysisinalkalineconditionswentfrom1.79×106_gmol−1_forcontrol AGPtoaslowas1.68×105_gmol−1_{forpapain-treatedAGP.Theoverallstructureoftheenzyme-treated} AGPswasfoundtobesurprisinglyquitesimilarwhatevertheenzymeusedandclose,withhowever somesubtledifferences,toAGunit.Atri-axialellipsoidconformationwasfoundinenzyme-treatedAGPs andthetwomainpreferentialdistancesidentifiedinthepairdistancedistributionfunctionwouldclaim infavorofrod-likeorelongatedparticlesoralternativelywouldindicatethepresenceoftwoparticles differingindimensions.Thesecondarystructurescontentofcontrolandenzyme-treatedAGPswere similar,highlightingboththehighrigidityoftheproteinbackboneandtheoverallsymmetryofAGP. ThisconclusionwasreinforcedbythemorecompactstructuresfoundwhenAGPwasintactcompareto themoreelongatedstructuresfoundwhenAGPwasenzymaticallycleaved.

Finally,thestructuralsimilaritiesfoundinenzyme-treatedAGPtogetherwiththetheoretical calcula-tionstoanalyticallyprobethetypeofbranchingwouldsuggestthatAGPwouldbemadeofaself-similar assemblyoftwotypesofbuildingblocks,thesecondbeingafive-foldrepetitionofthefirstone,for whichpalindromicaminoacidsequencewouldensureaself-orderingofcarbohydratemoietiesalong thepolypeptidechains.Thecleavagewouldthereforeleadtohydrolysedbuildingblockswithsimilar secondarystructuresandconformationswhatevertheenzymeused.

©2014ElsevierLtd.Allrightsreserved.

∗ Correspondingauthor.Tel.:+33240675052;fax:+33240675025. E-mailaddresses:drenard@nantes.inra.fr,denis.renard@nantes.inra.fr

(D.Renard).

1. Introduction

AGPsarethemost structurallycomplexhydroxyproline-rich proteinandareimplicatedtofunctioninplantgrowth, develop-ment,signaling,andplant–pathogeninteractions(Ellis,Egelund, http://dx.doi.org/10.1016/j.carbpol.2014.06.041

Schultz, & Bacic, 2010; Liu & Mehdy, 2007). They are found in plasma membranes, cell walls, and plant exudates (Seifert &Roberts,2007).AGPs aredefined by three criteria:the pres-enceoftypeIIarabino-3,6-galactanchains,ahydroxyproline-rich protein backbone, and the ability of most AGPs to bind to a classofsyntheticphenylazodyes, the␤-glycosylYarivreagents (Clarke, Gleeson, Harrison, & Knox, 1979). AGP protein back-bones are typically rich in hydroxyproline (Hyp) alternating with Ala, Thr, and Ser, whereas their carbohydrate moieties are mostly composed of galactose and arabinose with varying amounts of rhamnose, fucose, and glucuronic acid (Qu et al., 2008).The polysaccharide chains of AGPs are composed of ␤-(1,3)-galactan chains interrupted with ␤-(1,6)-galactopyranose (Gal) side chains and terminated mostly with arabinofuranose residues(Bacic, Churms,Stephen,Cohen, &Fincher, 1987;Tan, Qiu, Lamport,& Kieliszewski,2004; Tan etal., 2010).Recently, Tryfona,Liang,Kotake,Tsumuraya,Stephens,andDupree(2012) reported that Arabidopsis AGPs are similarly decorated with a linear ␤-(1,3)-galactan backbone with ␤-(1,6)-d-galactan side chains.

AGPfromAcaciagum,anexudateproducedbymanytreesand shrubsasa naturaldefense mechanism,wasalsodefinedasan arabinogalactan-proteinasmacromoleculesalsofulfilledthethree criteria classifying AGPs (Churms, Merrifield, &Stephen, 1983; Connolly,Fenyo,&Vandevelde, 1987a,1987b).Acaciagumis a highly heterogeneous material defined, on a hydrophobic sep-aration criterium,by three molecularfractions identified as an arabinogalactan-peptide (AG),an arabinogalactan-protein(AGP) and a glycoprotein (GP) fraction (Randall, Phillips, & Williams, 1989;Renard,Lavenant-Gourgeon,Ralet,&Sanchez,2006). Analy-sisofthefractionsshowedthateachcontainedsimilarproportions ofthevarioussugarsanddifferedessentiallyintheirmolecular massesandproteincontents.ThetertiarystructureofAGPhasbeen describedinliteratureintermsofawattle-blossom macromolec-ularassemblybyvirtueofwhichfew(_∼5)discretepolysaccharide blocksofMw ∼2×105gmol−1 areheldtogetherbyashort

pep-tidebackbonechain(Fincher,Stone,&Clarke,1983).Morerecently, Mahendran,Williams,Phillips,Al-Assaf,andBaldwin(2008) iden-tified smaller carbohydrate blocks of ∼4.5×104_gmol−1 _linked

by O-serine and O-hydroxyproline residues to the polypeptide chainofapproximately250residuesinlength.Thisresult,using achemicaltreatmentassumedtohydrolyzeonlythepolypeptidic backbone of AGP, contradicted the previous enzyme degrada-tion studies, using pronase to degrade polypeptidic backbone, whichconcludedthatthecarbohydrateblockswereoftheorder of∼(1.8–4.5)×105_gmol−1 _{(Aoki,Al-Assaf,Katayama,&Phillips,}

2007;Beltranetal.,2005;Chikamai,Banks,Anderson,&Weiping, 1996;Churmsetal.,1983;Connolly,Fenyo,&Vandevelde,1988; Duvallet,Fenyo,&Vandevelde,1989;Fincheretal.,1983;Osman, Menzies,Williams,Phillips,&Baldwin,1993;Randalletal.,1989). These apparent contradictory results coming from protease or chemical treatments on AGP could mean that the enzymatic kinetics conditions (e.g. pH, T, enzyme:substrate ratio, time of degradation)wouldberesponsibleforthedifferenceof molecu-larweightsfoundbetweenchemicalandenzymatictreatment.A closerexaminationof theenzymatickineticsconditionsin pre-vious literature was in favor of a complete hydrolysis of the proteinin theAGPcomponentafter24hofpronaseincubation (Aoki et al., 2007; Flindt, Al-Assaf, Phillips, & Williams, 2005; Mahendranetal.,2008).Pronaseisamixtureofseveralproteases withunspecificendogenousandexogenouscleavageofproteins, whichgenerallyproceedtothelevelofsingleaminoacids.This isnotthecasewhenchemicaltreatmentisused,suchassodium hydroxide/sodiumborohydrideusedbyMahendranetal.(2008), wherethechemical compoundsspecificallycleaveO-glycans at polysaccharide-polypeptidelinkagesinvolvingSerorThrresidues

butnotthoseinvolvingHypresidue.WhythusAGPchemical treat-mentwhichismorespecificthanpronaseforpeptidiccleavagescan leadtolowermolecularweightcomponents?Onestrong explana-tiontothis apparentcontradictionwouldcomefromtheworks of Aoki et al. (2007) and Al-Assaf,Sakata, McKenna,Aoki, and Phillips (2009)who clearlydemonstrated thatAGP isnot com-posedofasinglepolypeptidechainwithvariouscarbohydrateunits attachedtoit.TheAGPwould bemainlycomposedof arabino-galactan(AG) unitsassociated through hydrophobicassociation andthisiswhytheproteinsequenceandsugarcontentsis simi-larinbothfractionsbutdifferenttothatintheglycoprotein(GP) fractionasreportedpreviously(Mahendranetal.,2008;Renard et al.,2006;Williams, Phillips, &Stephen,1990).However,key to thestability and coherent of the AGP structure would be a smallproportion oflow molecularweighthighlyproteinaceous components oftenreported in theliterature as theGP fraction (Randalletal.,1989).Thisinterpretationwouldreconcile chemi-calandenzymaticdegradationwherethepresenceofpolypeptide backbones withdifferent lengths and as a consequence differ-ent carbohydrate blocks attachedto it would lead to different molecularweightcomponentsaftertreatmentwhateverthe con-ditions used. Al-Assaf et al. (2009) provided evidence that the highmolecularweightfraction(AGP)ingumarabicwasindeed highlyassociatedmolecularstructure.Inaddition,whensubjected tophysicaltreatments,suchasfiltrationorhighpressure homog-enization,AGPcanbedissociatedtoyieldthebasicAGmolecular unitswhichmakeupthegumandalowmolecularweightfraction highlyproteinaceous.Furtherevidencetosupportthis proposal camefromenzymedigestionofgumarabicwhichwasshownto bedependentonthestartingmolecularweight.Thereductionof themolecularweightfollowingenzymedigestionwasaslowas 1.8×105_or5.0×105_gmol−1_{whenmolecularweightofA.senegal}

gumprobedbyHPSEC-MALLSwasprocessedasonepeak(Al-Assaf etal.,2009).Finally,therecentresultsreportedbyMahendranetal. (2008)clearlyestablishedbyelectrophoresisafterdeglycosylation ofgumarabicthepresenceoftwopolypeptidicchains,one corre-spondingtoacoreproteinof∼250aminoacidresiduesthatwas consistentwiththeliteratureforAGPcoreproteins(Williams& Langdon,1995)andasecondputativecoreproteinof45aminoacid residuesthatwouldbeconsistentwiththeproteinaceous compo-nentassociatedtotheAGgumfraction(Mahendranetal.,2008). WeidentifiedrecentlytwotypesofconformationsforAGPin solu-tion,alowmolarmasspopulationwithlong-chainbranchingand acompactconformationandahighmolarmasspopulationwith short-chain branching and an elongated conformation(Renard, Garnier,Lapp,Schmitt,&Sanchez,2012).Themaximumdimension ofAGPwasfoundtobe64nmandelectronmicroscopy observa-tionsclearlyrevealedthepresenceofinteractingbranchedchains, aggregatechainsandring-likestructuresthatrenderedthe identifi-cationofsingleAGPmacromoleculesverydifficult(Renard,Garnier, Lapp,Schmitt, &Sanchez,2013).Thesestructuralstudieswould lead tothemain conclusions that AGP couldbean association of atleast two differentunitsand that thesemolecular associ-ations would be pronetoaggregation when adsorbed onsolid surfaces.

The main objectives of the present study were to probe thestructure ofthe AGPfraction fromAcaciagum in solution, after purification by hydrophobic interaction chromatography, and enzymatic degradationusing severalproteases,in order to try to reveal new insights into the structure of this complex assembly. The combination of high-performance size exclusion chromatographycoupledtoon-linemulti-anglelaserlight scat-tering(HPSEC-MALLS),smallangleneutronscattering(SANS)and circular dichroism(CD) measurementswas performedin order to reveal some fundamental knowledge of the carbohydrate-polypeptidelinkagesoccurringinAGP.

2. Materialsandmethods

2.1. Materials

Spray-driedAcaciagum(lot97J716)fromA.senegaltreeswasa giftfromCNIcompany(Rouen,France).Beforepurification,Acacia gumwasdialyzedagainstdeionizedwaterandthenfreeze-dried. Allchemicalswereof analytical gradeand wereobtainedfrom Merck,SigmaandAldrich.

2.2. PurificationofAGPmolecularfraction

TheAcaciagumarabinogalactan-protein(AGP)molecular frac-tion was purified by hydrophobic interaction chromatography (HIC),aspreviouslydescribed(Renardetal.,2006).PurifiedAGP wasdialysedandthenfreeze-driedbeforefurthercharacterization. 2.3. Acidicandalkalineenzymehydrolyses

Bothacidic and alkalineenzymatic treatments wereapplied onAGP(andAG)molecularfractionfromAcaciagum.Foracidic hydrolyses, AGP and enzymes were dissolved in 0.1M acetate bufferpH4.EnzymesusedwereendoproteinaseGlu-Cfrom Staphy-lococcusaureusstrainV8(lot1278409,FlukaBiochemika),pepsin from porcine gastric mucosa (lot 035K76701, Sigma–Aldrich) and phosphatase acid type IV-S from potato (lot 026K7014, Sigma–Aldrich).Foralkalinehydrolyses,AGPandenzymeswere dissolvedin 0.1Mphosphate buffer pH8. Enzymes used were subtilisinA, type VIII fromBacillus licheniformis (lot 016K1295, Sigma–Aldrich),pronasefromStreptomycesgriseus(lot084K1843, Sigma–Aldrich),trypsin,TPCKtreated,frombovinepancreas(lot 026K7770,Sigma–Aldrich),papainfromCaricapapaya(lot108014, BoehringerMannheim)andproteinaseKfromTritirachiumalbum (lotP-6556,Sigma–Aldrich).Forallassays,3mgofsubstrate(AGor AGP)wasdissolvedin0.3mLoftheappropriatebufferand0.3mL ofenzymeataconcentrationof0.2mg/mLwasadded,givingafinal substrateconcentrationof5mg/mLandasubstrate:enzymeratioof 25.HydrolyseswereperformedonenightatT=37◦C.Controlswere performedusinguntreatedAGandAGP.Experimentswere tripli-catedandsampleswerethenstoredat4◦Corextensivelydialysed againstdistilledwaterandfreeze-driedbeforefurtheranalyses. 2.4. HPSEC-MALLS

Thedeterminationofmolecularweightandsizedistributionsof AGP(andAG)afterenzymatichydrolyseswereperformedby cou-plingonlineahigh-performancesizeexclusionchromatography toamulti-anglelaserlightscatteringdetector,adifferential refrac-tometerandadifferentialviscosimeter.AGP(orAG)dispersions at0.05wt%storedat4◦Candcontainingenzymesorsolubilized afterfreeze-dryingat0.04wt%into50mMNaNO3buffercontaining

0.02%NaN3aspreservativewerefilteredthrough0.2␮mAnotop

membranes(Anotop,Alltech,France),andinjectedonaHPSEC sys-temconstitutedofaShodexOHSB-Gpre-columnfollowedbytwo ShodexOH-pack804HQand805HQcolumnsusedinseries.The sampleswereelutedat0.7mlmin−1witha50mMNaNO3solution.

On-linemolecularweightand intrinsicviscositydeterminations wereperformedatroomtemperatureusing amulti-angle laser lightscattering(MALLS)detector(mini-Dawn,Wyatt,Santa Bar-bara,CA,operatingatthreeangles:41,90and138◦),adifferential refractometer(ERC7547A)(dn/dc=0.153mLg−1)anda differen-tialviscosimeter(T-50A,Viscotek).Weight(Mw)andnumber(Mn)

averagemolarmass(gmol−1),andradiusofgyration(Rg,nm)were

calculatedusingAstra1.4software(Wyatt).Intrinsicviscosity[_] wascalculatedusingTriSECsoftware,version3.0(Viscotek).No dif-ferenceswereobservedintheresultswhateverthephysicalstate

andthepresenceornotofenzymeinthesamples(hydrated sam-plescontainingenzymesvs.solidsampleswithoutenzymes). 2.5. SANSexperimentsanddatatreatment

SANS experiments and data treatment were performed as described previously (Sanchez, Schmitt, Kolodziejczyk, Lapp, Gaillard, & Renard, 2008). Briefly, scattering experiments on filtratedAGP(intactorenzymaticallyhydrolysed)(C=1wt%) dis-solved in 50mM NaCl solution in 99.8% D2O in order to get

theformfactorP(q)wereperformedonthePAXYinstrumentat the Laboratoire Léon Brillouin (LLB) (Saclay, France) using two spectrometerconfigurations:=8 ˚A(incidentwavelength),d=2m (distanceofthesampletothedetector)and=8 ˚Aandd=6.74m. Thewave-vectorq=4/ sin(/2), beingthescatteringangle, rangedbetween4×10−2and3nm−1.Thedatawerenormalized fortransmissionandsamplepathlengthanddividedbythewater spectrum.Anabsoluteintensityscalewasobtainedusingthe abso-lutevalueof waterintensityinunitsof cross-section(cm−1).A subtractionoftheappropriateincoherentbackground,takingH/D exchangeonAGPintoaccount,wasrealized.TheformfactorP(q)of intactorcleavedAGPwasobtainedat50mMNaClconcentration whereintermolecularinteractionswerenegligibleandtherefore S(q)=1.

The pair distance distribution function P(r) together with the structural parameter derived from P(r), i.e. the maximum dimensionoftheparticle(Dmax), wascomputedfromtheform

factorbytheindirectFouriertransformprogramGNOM(Svergun, 1992; Svergun, Semenyuk, & Feigin, 1988). Fit of experimen-tal AGP form factor by different theoretical form factors was assayedusingtheSansViewsoftwarev2.0.1(http://danse.chem. utk.edu/sansview.html).

2.6. FarUV-CD

ThecirculardichroicspectraofintactorcleavedAGPin0.1M phosphatebufferpH8wererecordedonaCD6Dichrograph Instru-ment(Jobin-Yvon,Longjumeau,France)at25◦C,calibratedusing ammoniumd10-camphorsulfonicacid.Spectrawereaveragedover

twoscanswithabandwidthof1nmandstepresolutionof0.5nm. Allspectrawerereportedintermsofthetamachineunits(mdeg) withinthe185–260nmspectral areausinga1-mmpath-length cylindrical cuvette. AGPconcentrations used(after filtration on 0.2␮mporesizes)wereof0.5wt%.Allspectrawerecorrectedfor thesolvent(buffer+enzyme);noisereductionwasapplied accord-ingtotheJobin-Yvonsoftware.CDspectraacquisitionwasrepeated twice.

CD spectra were analyzed in terms of secondary structure content using Selcon3 program (Sreerama & Woody, 1993) in Dichroweb(Lobley, Whitmore, &Wallace, 2002) and usingthe proteindataset3(185–240nm,37proteins),6(185–240nm,17 proteins) and7 (190–240nm, 42 proteins)as thereferenceset (Whitmore&Wallace,2004,2008).

3. Results

Theenzymatichydrolysesofthetwomainmolecularfractions, AGandAGP,fromA.senegalgum,wereperformedinacidic(pH 4)andalkaline(pH8)conditionsusingenzymesinthe appropri-atepH-optimaconditions.Table1summarizedthemodeofaction ofeachenzymeusedin0.1MacetatebufferpH4and0.1M phos-phatebufferpH8conditions.Theenzymaticconditionsusedin thisstudythuscoveredawiderangeofpossibilityforpeptideand esterbondscleavageinthepolypeptidicbackboneofAGandAGP unlessthepolypeptidicchainswereaccessibletoenzymesinthe

Table1

Modeofactionoftheenzymesusedtocleavepolypeptidicbackbonesofarabinogalactan(AG)andarabinogalactan-protein(AGP)molecularfractionsfromAcaciagum.

Bufferconditions Enzyme Specificitya _{ProportioninAGP(%)}b

0.1Macetatebuffer,pH4 EndoproteinaseGlu-C Glu,Asp-protease 6.5

0.1Macetatebuffer,pH4 Pepsin Phe,Tyr,Trpprotease 4.4

0.1Macetatebuffer,pH4 Phosphataseacid Esterase(oneesterbondandaphosphategroup) –c

0.1Mphosphatebuffer,pH8 SubtilisinA Serineprotease 12.2

0.1Mphosphatebuffer,pH8 Pronase Mixtureofseveralproteases 100

0.1Mphosphatebuffer,pH8 Trypsin Arg,Lysprotease 3.6

0.1Mphosphatebuffer,pH8 Papain Arg,Lys,Glu,His,Gly,Tyrprotease 20.7

0.1Mphosphatebuffer,pH8 ProteinaseK Aliphatic,aromatic,hydrophobicprotease 24

a_{Glu:glutamicacid;Asp:asparticacid;Phe:phenylalanine;Met:methionine;Leu:leucine;Trp:tryptophan;Arg:arginine;Lys:lysine.Aromatic:Phe,Tyr,Trp.Hydrophobic:}

Ile,Leu,Val,Cys,Met.Aliphatic:Gly,Ala.

b_{FromRenardetal.(2006).} c _Unknown.

twofractions.Aliquotsoftreatedsampleswerethenanalyzedby HPSEC-MALLS,SANSandfarUV-CDasdescribedabove.

3.1. HPSEC-MALLSmeasurements

Itwasfirstobservedthatacidicandalkalineenzymatic treat-mentsdidnot affectthestructure of AGwhatever theenzyme used (data not shown). HPSEC-MALLS results for AG before andafterenzymatic hydrolysesgave amolecularweightMw of

2.82×105_gmol−1_{, a polydispersity index Mw/Mn of 1.26and a}

radiusofgyrationRgof8±5nm,valuesincloseagreementwith

thosereportedpreviously(Sanchezetal.,2008).Thisresultwould mean thatthe peptidechain of AGwould not be accessibleto enzymewhateverthepHconditionsandenzymesused.Fromour previousresultandthose obtainedby Mahendranetal.(2008), theprotein corein theAGcomponent would consist of43 (or 45)aminoacids(Mahendranetal.,2008;Renardetal.,2006).The proteincorewouldbehowevercompletelyburiedbythe oligo-andpolysaccharideblocksattachedtohydroxyproline(Hyp), ser-ine(Ser) orthreonine(Thr)residues,andasaconsequence,not accessibletoenzymes.Indeed,arabinoseandrhamnosewerefound tobeattheperipheryofthemacromolecules(Anderson,Hirst,& Stoddart,1966;Mahendranetal.,2008).Moreover,itwasclearly establishedbyMahendranetal.(2008)thatthechemicaltreatment usedtodegradeproteinaceouscomponentsinAcaciagumdidnot affecttheprotein-deficientAGreinforcingthefactthat oligosac-charidesandpolysaccharideblockswerepreferentiallylocatedat theperipheryofthemacromolecule.Inaddition,itwasalso inter-estingtonotethatWilliamsandPhillips(2000)demonstratedthat thethirdminorglycoprotein(GP)fractionwasunaffectedbya pro-teasetreatment.Thequestionforwhichweshouldanswerthrough thisstudyisthefollowing:ifAGandGPfractionsareunaffectedby proteasetreatments,atleastinalkalineconditions,andifAGPis amolecularassociationofAGandGPunits(Al-Assafetal.,2009), whichwillbethereactivityofAGPtowardproteasetreatment?

ThesecondresultsweobtainedwerethatAGPwasalso unaf-fectedbyacidicenzymatictreatmentwhatevertheenzymeused. The molecular weight Mw and theradius of gyration Rg were

1.83×106_gmol−1 _{and 33nm, respectively,values very closeto}

thosereportedpreviouslyfornativeAGP(Renardetal.,2006,2012). Theabsenceof reactivity ofenzymesused inacidic conditions, endoproteinaseGlu-C,pepsinandphosphataseacid,wouldmean thattheenzymeswereunspecifictowardaminoacidresiduesin thepolypeptidechainsofAGPorwereunabletoproceeddueto sterichindrance.Thefirsthypothesiswasnottruesinceallamino acidresiduescleavedbythethreeenzymeswerepresentinAGP (Renardetal.,2006).Thesecondhypothesiswouldthereforebe moreconceivable,aspreviouslydiscussedfortheAGfractionfrom Acaciagum.However,allpreviousenzymaticstudiesonAGPand wholegumperformedinalkalineconditionsdemonstratedpartial

10 12 14 16 18 20 0.0 5.0x10-5 1.0x10-4 1.5x10-4 V e (mL) C (g /mL )

Fig.1. HPSECchromatogramshowingtheelutionprofilemonitoredbyrefractive index(C,g/mL)andlightscattering(Mw,gmol−1)forAcaciagum

arabinogalactan-protein(AGP)beforeandafterenzymatichydrolysesasafunctionofelutionvolume (Ve,mL)obtainedfromHPSECcoupledtoon-lineviscosimeterandmulti-anglelight

scattering (MALLS)detector.HPSECchromatogramforarabinogalactan-peptide (AG)molecularfractionisdisplayedforcomparison(continuousline).AGP( ), AGP+proteinaseK( ),AGP+subtilisine( ),AGP+pronase( ),AGP+papain ( ).

degradationofAGP.Thesterichindrancewouldthusnotbethe mainreasontotheabsenceofenzymesreactivityinacidic condi-tions.Onepossibleexplanationtounderstandtheseresultswould bethatthelowcontentofaminoacidresiduessusceptibletobe cleaved byenzymes (Asp,Glu,Phe, Tyr,Trp residues), residues thatinadditioncouldalsobeburiedandinaccessibletoenzyme, wouldnotaffectthewholestructureofAGP.Thesecond proba-bleexplanationwouldbethatnoncovalentinteractionsbetween sugarmoietiesand/orsugarmoieties/aminoacidresidueswould takeplacemaintainingtheoverallstructure ofAGPevenin the presenceofenzymecleavages.

AGPwasthereforeaffectedbyenzymatictreatmentsin alka-lineconditionsasillustratedinFig.1whichdisplayedtheelution profilesofAGPobtainedbyHPSEC-MALLSafteronenightof enzy-matic treatment at T=37◦C. HPSEC chromatograms for control arabinogalactan-peptide(AG)andarabinogalactan-protein(AGP) weredisplayed forcomparison. It wasfirstevidenced fromthe maximaintheconcentrationpeakasafunctionofelutionvolume thatAGPwasdegradedtosomeextentdependingontheenzyme used. In addition, one single peak was observed whatever the enzymeusedwithhoweverahigherpolydispersityindexMw/Mn

valuecomparetocontrolAGP,indicatingthatAGPmacromolecules wouldbeuniformlycleavedinsolutiongivingrisetohomogeneous fragments.WhilecontrolAGPhada Mw of1.79×106gmol−1,a

Table2

Summaryofarabinogalactan-protein(AGP)fromAcaciagum,beforeandafterenzymatichydrolysesusingproteinaseK,subtilisin,pronaseandpapain,structuralparameters determinedbyHPSEC-MALLSandSANS,orcalculated.

Structuralparameters AGP AGP+pronase AGP+proteinaseK AGP+subtilisin AGP+papain

HPSEC-MALLS Non-hydrolysedAGPa _{100 23.5 11.4 5.8 0} Mw(gmol−1) 1.79×105 6.0×105 4.9×105 3.67×105 1.68×105 Mw/Mn 1.32 1.43 1.47 1.75 1.46 Rg(nm)b 33 21.2 18 17.7 11 [](mLg−1) 73.5 25.4 26.3 18.2 15.4 Rh(nm)c 27.5 13.4 12.7 10.2 7.4 (Rg/Rh) 1.2 1.6 1.4 1.7 1.5 SANS P(0)(cm−1_{) 3.3 1.77 1.52 1.25 1.22} Rg(nm)d 18.4 14.8 14.4 14.5 16.2 Rg(nm)e 18.5 14.7 14.6 15.6 16.8 Dmax(nm)e 64 49.4 49.5 51.5 54.2 Ra(nm)f 0.3 0.4 0.5 0.3 2 Rb(nm)f 9.3 7.1 6.5 5.8 4.2 Rc(nm)f 32 27 26.5 23 27 2g _{2.67 0.84 0.85 0.72 2.26}

a_{Estimationofnon-hydrolysedAGPfromHPSEC-MALLS(seetextfordetails).} b _{FromSEC-MALLSbylightscattering.}

c _{Calculatedfromtheequation[]=6.308×10}24_R h3/Mw. d _{FromtheGuinierregionoftheformfactorP(q).} e_{FromthepairdistancedistributionfunctionP(r).} f _R

a,Rb,Rc,semi-axisofthetriaxialellipsoidmodelthatbestfitstheformfactorsP(q). g_{Parameterofbestfitquality.}

trypsintreatment(datanotshown)forwhichMwwasunchanged

(Table2).AmolecularweightMwdecreasefrom1.79×106toas

lowas1.68×105_gmol−1 _{wasobservedfor papain-treatedAGP.}

Moreover,theintrinsic viscosity[]value decreased from73.5 to15.4mLg−1.Aroughestimationoftheenzymeefficiencywas donebycalculatingthepercentageofnon-hydrolysedAGPafter onenightatT=37◦C.Forthat,itwasconsideredthatintactAGP wasstill present after enzymatic hydrolysis if macromolecules with Mw higher than ∼1×106gmol−1 (i.e. Ve∼13.5mL) were

present.Theratioofthesumofconcentrationsformacromolecules eluted for Mw>1×106gmol−1 over total concentrations given

bythe refractometer gave theresults inTable 2 where papain appeared to be the most efficient enzyme with 100% hydrol-ysedAGPwhilepronasewasfoundtobetheless efficientwith 23.5%residualintactAGPafteronenightofenzymaticreaction. These rough estimations of enzymatic efficiencies were how-everinaccordancewiththelowestMwfoundforpapain-treated

AGP,Mw=1.68×105gmol−1,whilethehighestonewasfoundfor

pronase-treatedAGP,Mw=6.0×105gmol−1.

Theseresultswereinpartialagreementwithprevious litera-turedatastatingthatpronasetreatment,fortimeofreactionequal orhigherthanonenightandenzyme:substrateratioof56or60, leadtoMw∼2–3×105gmol−1and[]∼11–12mLg−1,valuesclose

tothose ofAGunit(Al-Assafetal.,2009;Connollyetal.,1988; Elmanan,Al-Assaf,Phillips,&Williams,2008).ThehigherMwfound

inourstudywasclearlyexplainedbythelowerenzyme:substrate ratioused.Theabsenceofcleavageofpeptidicbondsusingtrypsin couldberelated,asforenzymesinacidicconditions,tothevery low content of arginine and lysine residues in AGP (Table 1), residuesthat in addition couldalso beburied and inaccessible toenzyme(Renardetal.,2006).Ananalysisoftheconformation adoptedbyAGPbeforeandafter enzymatictreatmentrevealed thatenzyme-treated AGPwasvery similarto AGconformation contrarytocontrolAGPthatadoptedtwodifferentconformations (Fig.2).AGPwasthereforefoundtoadoptacompactconformation forthelow molarmasspopulationandanelongated conforma-tionforthehighmolarmasspopulation(Renardetal.,2012).The HPSECchromatogramsshowingtheMwdependencewiththe

elu-tionvolumeVeclearlyrevealedthegreatsimilarityfoundbetween

controlAG and hydrolyzedAGP,highlighting thehypothesis of

11 12 13 14 15 16 17

10000 100000

1000000 Unique conformation to AGP

similar conformation to AG and hydrolysed AGP

Mw

(g/

m

ol)

V_e (mL) similar conformation to

AG, AGP and hydrolysed AGP

Fig.2.HPSECchromatogramshowingthemolecularweight(Mw,gmol−1)

depend-encewiththeelutionvolume(Ve,mL)forAcaciagumarabinogalactan-protein

(AGP)beforeandafterenzymatichydrolyses.Molecularweightdependencefor arabinogalactan-peptide(AG)molecularfractionisdisplayedforcomparison (con-tinuousline). AGP ( ), AGP+proteinase K ( ), AGP+subtilisine( ), AGP+pronase( ),AGP+papain( ).

thedissociationofAGPintobuildingblocks similartoAGunits afterenzymatictreatment.DifferencesinMwofhydrolyzedAGP

evidencedseveraldegreeof hydrolysisaccordingtoenzyme,as demonstratedbythecontentofthenon-hydrolysedAGP(Table2), andhighlightedtheexistenceofa buildingblockwithlowMw,

i.e.1.68×105_gmol−1 _{forpapain-treatedAGP.Theassociationof}

thesebuildingblocks leadingtoAGP macromolecularassembly couldoriginatefromcovalentlinkagesand/orfromhydrophobic associationsandhydrogenbonding,covalentlinkagesbeinga rea-sonablehypothesisconsideringthemodeofactionofproteases usedinthisstudyandthemixedcompositionofAGPcontaining glycomodulesofbothextensinandarabinogalactan-glycoproteins (Goodrum, Patel, Leykam, & Kieliszewski, 2000). In particular, disulfide intermolecular bridges, neither however identified in AGP, and/or isodityrosine intramolecular cross-links, found in extensins(Kieliszewski&Lamport,1994),couldbepresentinAGP

10-1 100 10-3 10-2 10-1 100 P( q) (cm -1 ) q (nm-1) -2.4 -3.8 -2.2 -1.1

Fig.3.FormfactorsobtainedbySANSat25◦_{Cand50mMNaClon}

arabinogalactan-protein(AGP)dispersions(C=1wt%)fromAcaciagumbeforeandafterenzymatic hydrolyses.(AG)molecularfractionisdisplayedforcomparison(continuousline). AGP( ),AGP+proteinaseK( ),AGP+subtilisine( ),AGP+pronase( ), AGP+papain( ).Formfactoronarabinogalactan-peptide(AG)( )isdisplayedfor comparison.Highqrangeofformfactorscanbedescribedbypowerlawexponents whichrangeisindicatedinthefigure(−3.85forAGPand2.4forAGP+pronase). Intheintermediateqrange,powerlawexponentswerecomprisedbetween2.13 (AGP)and2.2(AGP+papain).

macromolecularassemblygivingabetterefficiencytoward hydrol-ysistopapain.Theuniqueglobularconformationhoweveradopted bycontrolAGPforVehigherthan13mLcomparedtocontrolAGand

enzyme-treatedAGPwasinaccordancewithourpreviousresults statingthat AGPwould behavein solutionasa branched poly-merwithconformationsrangingfromglobulartoelongatedshape dependingontheMwrangeconsidered(i.e.sizeofthecarbohydrate

branches)(Renardetal.,2012).Fromthesepreliminaryresults,it wasconcludedthatenzyme-treatedAGPsadoptedonesingle con-formationinsolutionclosetotheAGunitwhateverthemolecular weight,contrarytocontrolAGPthatadoptedtwodistinct confor-mationsdependingonthemolecularweight.Theapparentsimilar conformationadoptedbyenzymetreatedAGPswithAGunitshould thereforebeshadedbythefactthatlowerMwcomparetoAGwere

obtainedforhydrolysedAGPs.Thisimportantresultwould there-foremeanthat,structuralblocks,differentfromAGunits,could beattheoriginof themacromolecularassembly leadingtothe complexAGPmacromolecule.

3.2. SANSmeasurements

Additionalinformationsonconformationand structurewere probedusingsmallangleneutronscatteringexperimentsona dis-tanced=2/qranging from2to∼150nm. FormfactorsP(q)of enzyme-treatedAGPand controlAGPin 50mMNaClasa func-tionofthewavevectorq(nm−1)weredisplayedinFig.3.Atlow scatteringwavevectorsq,thescatteringfunctionP(q)vs.qof con-trolAGPfollowedapowerlawwithanexponentof1.1,indicative ofa linearconformationtypicalof arodoranelongated cylin-der,asdescribedpreviously(Renardetal.,2012).Enzyme-treated AGPsdisplayedlowerscatteringintensityatlowqvalues indica-tiveoflowerMwcomparetocontrolAGP,inagreementwithMw

determinedbyHPSEC-MALLS.Thisdecreasingmolecularmasswith enzymetreatmentwasfurtherconfirmedthroughtheradiiof gyra-tion Rg values that decreased from 18.4nm for control AGPto

14.4nmforproteinaseK-treatedAGP.Inaddition,itwasnoticed that Rg values werequite constant whatever theenzyme used

withhoweveranintermediatevaluefoundforpapain-treatedAGP.

Thesepreliminaryresultswouldfavor againaconsensusin the wholeconformationof enzyme-treatedAGP whatever thetype ofenzyme used.Atintermediatescattering wavevectorsq,the scatteringfunctionofcontrolAGPandenzyme-treatedAGPwere superimposedandfollowedapowerlawwithanexponentvalue of2.2.Avalueof2isindicativeofthepresenceofanisolated pop-ulationofdisk-likeparticles,andmoregenerallyofa2Dparticles morphology(Svergun&Koch,2003),orofafractalstructure.The affinityof controlAGPandenzyme-treatedAGPfor thesolvent wasthusestimatedtobeof0.45(1/exponentvalue),value corre-spondingtoathetasolvent,intermediatebetweenapooraffinity (0.33)andavery goodaffinity(0.6)for thesolvent.Inthehigh qrange,wherethelocalstructureofscatteringobjectsisprobed, scatteringintensityfollowedapowerlawwithanexponent ran-gingfrom3.8forcontrolAGPto2.4forcontrolAGandpapainand pronase-treatedAGPs.Theseexponentvaluesclearlyindicatedthat controlAGPhadafractalsurfacestructurewhileenzyme-treated AGPsandcontrolAGhadaninternalstructureratherdenseand ramified(Renardetal.,2012;Sanchezetal.,2008).Itwas interest-ingtonoticethatthescatteringintensityI(q)vs.qofcontrolAG wassuperimposedtoenzyme-treatedAGPsonaqrangeranging from_∼0.3to_∼0.8nm−1,coveringadistancebetween2.5and20nm (Fig.3).Thiswasparticularlytrueforpapainandpronase-treated AGPs,slightdeviationsbeingobservedathighqforproteinaseK andsubtilisin-treatedAGPs.Theslightdifferencesobservedatlow qvalueshighlightedthestructuralchangesoccurringafter enzy-matictreatmentonAGPmacromoleculescomparetocontrolAG. ThestructuraldifferencesidentifiedbySANSandnotby HPSEC-MALLSwhencontrolAGwascomparedtoenzyme-treatedAGPs wererelatedtothedistanceprobedusingneutronscattering com-paretolightscattering.Thehigherdistancesthereforeprobedby lightscatteringmaskedsomesubtlestructuraldifferencesobserved byneutronscatteringatlowerdistances.Theseresultswould rein-forcethehypothesisoftheassociationofstructuralbuildingblocks differentfromAGunitsleadingtotheformationofAGP macro-molecularassembly.

Since thearabinogalactan-peptide (AG)and arabinogalactan-protein(AGP)fractionsweredescribedasathinoblateellipsoid (Sanchezetal.,2008)andatri-axialellipsoidoranellipticalcylinder (Renardetal.,2012),respectively,afitoftheenzyme-treatedAGPs formfactorswasattemptedtakingintoaccountthefactthat HPSEC-MALLSrevealedelongatedconformationsand SANS identifieda rather elongatedstructure andparticles with2Dmorphologies. Finally,itappearedthatatri-axialellipsoidformfactorreasonably fittedexperimentaldata(Fig.4).Detailsofthesemi-axesvaluesof thetri-axialellipsoidsweregiveninTable2.Thefitswerequite sat-isfyingandreasonablyapproachedtheoverallcharacteristicofthe formfactor.Itwasthereforesurprisingtonoticethat,evenifa10 foldreductionofmolecularmassoccurredafterpapaintreatment onAGPforinstance,theoverallstructureofthemacromoleculewas maintained.Thestructuralsimilaritiesobservedbeforeandafter enzymatictreatment,leadingtoatri-axialellipsoidmorphology, evenifsubtledifferenceswereprobedatsmalldistances(i.e.high qvaluesbySANS),couldberelatedtothelackofenzymespecificity and/orthesterichindrance,duetothepresenceoflarge carbohy-drateblocksattachedtothepolypeptidebackbone,avoidingthe enzymestofreelydiffuseinthecoreoftheAGP.Theconsensusin themodeofenzymeaction,due totherepetitionofaminoacid sequencesinAGPandalsosterichindrance,wouldthereforelead tothesamestructuralblocksafterdegradationwhateverthetype ofenzymeused.

Additionalinformationaboutthestructuralmorphologyofthe controlandenzyme-treatedAGPscame fromtheP(r)functions, shown inFig.5, obtainedbytheGNOM treatmentof theSANS data.Fromthisfunction,themaximumdistances(Dmax)foundfor

0 5 10 15 20 25 0 1 2 3 4 5 6 7 8 P( q) (qR g ) 2 qR_g

Fig.4.Kratky-typeplotsofformfactorsandcurve-fittingbyatri-axial ellip-soidmodelofformfactorsarabinogalactan-protein(AGP)dispersions(fromAcacia gum)beforeandafterenzymatichydrolyses.AGP( ),AGP+proteinaseK( ),AGP+subtilisine( ),AGP+pronase( ),AGP+papain( ).Formfactoron arabinogalactan-peptide(AG)( )isdisplayedforcomparison.

(Table2).Inaddition,Rgvaluescalculatedfromthedistance

distri-butionfunctionswereinverygoodagreementwiththeRgvalues

obtainedinthe Guinieranalysis(Table2).Note that these val-ueswereusedtofittheexperimentalformfactorsbythetri-axial ellipsoidtheoreticalformfactor.Theextended,tail-likedecayof theP(r)functionsindicatedelongatedAGPsthatvaryfrom glob-ular/ellipsoidtocylindrical(Glatter,1979,2002), conformations inagreementwiththosereportedfromHPSEC-MALLSdata.Two preferential distances, three distances in the case of enzyme-treatedAGPs,werethereforeidentifiedastwo,orthree,maxima wereobserved in the P(r) functions.The second maximawere commontocontrolandenzyme-treatedAGPsandcorresponded toa half-distanceof about32nm. The firstmaximawasfound tobe at a distanceof 10nm for control AGP while a distance of about7.5nmwas obtainedfor enzyme-treated AGPsexcept

0 10 20 30 40 50 60 70 0.0 0.2 0.4 0.6 0.8 P(r) (x10 -3 ) r (nm) D_max

Fig.5. PairdistancedistributionfunctionP(r)calculatedbyindirectFourier trans-formforthearabinogalactan-protein(AGP)fromAcaciagumformfactorusing GNOMsoftwarebeforeandafterenzymatichydrolyses.AGP( ),AGP+proteinase K( ),AGP+subtilisine( ),AGP+pronase( ),AGP+papain( ).Form fac-toronarabinogalactan-peptide(AG)( )isdisplayedforcomparison.Maximum dimensions(Dmax)indicatedbyarrowswerecomprisedbetween64nm(AGP)and

49.5nm(AGP+subtilisine).TheRgobtainedfromtheintegralsoftheprofileswere

comprisedbetween18.5nm(AGP)and14.6nm(AGP+proteinaseK)(seeTable2).

190 200 210 220 230 240 250 260 -4 -2 0 2 θ (m deg ) (* 1 0 -4 ) λ (nm)

Fig.6.Far-UVcirculardichroismspectraobtainedat25◦_{CandphosphatebufferpH}

8forarabinogalactan-protein(AGP)dispersionsfromAcaciagumbeforeandafter enzymatichydrolyses.AGP( ),AGP+proteinaseK( ),AGP+subtilisine( ),AGP+pronase( ),AGP+papain( ),()collagen.

papain-treatedAGPforwhichalowerdistanceof5nmwasfound. Note thatthefirst maximawereclosein shape and dimension to the control AG fraction, with however a larger asymmetric distributioninthecaseofenzyme-treatedAGPs.Thethird max-ima,onlyidentifiedinenzyme-treatedAGPs,correspondedtoa half-distanceofabout40nm.ThedistancesidentifiedintheP(r) functionswouldcorrespond,inthecaseofrod-likeorelongated particles,toacross-sectionaldimensionof10nm(or7.5or5nm forenzyme-treatedAGPs)andamajordimensionlengthof64nm forcontrolAGPand about50nmforenzyme-treatedAGPs.The cross-sectionaldimensionswerethereforeinverygoodagreement with the Rb semi-axis values calculated in the case of a

tri-axialellipsoidmodel(Table2).Alternatively,thepresenceoftwo orthreemaximaintheP(r)functionscanalsoindicatethe pres-enceoftwo orthreeparticlesdifferingindimensionsor,inthe particularcaseofproteins,distancesseparatingcovalentlylinked domainsinasingleproteinormonomersinamultimericprotein (Wangetal.,2009).IntheparticularcaseofAGP,the macromolecu-larassemblycouldcomefromtheassociationofcovalently-linked structuralbuildingblocks.

3.3. Far-UVcirculardichroism

ThemolecularscaleofAGPandenzyme-treatedAGPswas stud-iedusingfar-UVcirculardichroisminordertoprobethesecondary structuresof thepolypeptidebackboneafterprotease cleavage. Acollagenstandard wasalsodisplayed forcomparison (Fig.6). Aminimumat203nmanda maximumat225nmwere identi-fiedintheCDspectrainthe185–260nmspectralregionofAGP andenzyme-treatedAGPs.Collagen,anexampleofproteinwith a PPII helix conformation, displayed a strong negative bandat 200nmandaweakpositivebandat220nm(Fig.6).Aminimum at205nmand amaximumat226nmwerepreviouslyfoundin a polyhydroxyproline standard, corresponding toa polyproline II(PPII)conformation(Renardetal.,2006).ThePPII-helix struc-tureis generally foundon thesurfaces of proteinswhere it is abletohydrogenboundwiththesurroundingwater molecules (Sreerama &Woody, 1993). ThisPPII type secondary structure wouldbepartiallyadoptedbyAGPandenzyme-treatedAGPs.It is wellknown that thoughthe conformationalfreedom ofPPII canbepartiallylimitedwhenitincludesprolineduetoits pyrro-lidinering,theextendedPPIIhelixismarkedly moreflexiblein comparisonwiththe␣-helixand ␤-sheet(Adzhubei,Sternberg,

&Makarov,2013).CDsecondary structureanalysesof AGPand enzyme-treatedAGPs,usingtheself-consistentmethod(Sreerama &Woody,1993)inDichroweb(Lobleyetal.,2002),predicted sim-ilarsecondary structurescontent: 11%␣-helix,25 ␤-sheet,22% ␤-turn,26%PPIIhelixand16%unorderedstructure.Thepresenceof unorderedandPPIIsecondarystructuresisthebasiccharacteristic ofnativelyunfoldedproteins(Uversky,2013a,2013b),highlighting theextendedandflexibleconformationofthepolypeptide back-boneinAGPs.Theoreticalsecondary structurepredictionsmade onthe19-residuessequenceidentifiedbyGoodrumetal.(2000)in theAGPconsensusglycopeptide,usingalgorithmssuchasi-TASSER (http://zhanglab.ccmb.med.umich.edu/I-TASSER/),FoldIndex, Dis-EMBL, GLOBPLOT, SPRITZ, PONDR-FIT, DISPRO-Scratch protein predictor or PSIPRED (http://expasy.org/), gave however 100% unordered structure. The result was unchanged whatever the lengthoftheconsensusglycopeptide(secondarystructureswere alsopredictedfortheconsensussequencerepeatedseveraltimes intandem).Inaddition,PSIPREDalgorithmfoundahigh flexibil-ity in theconsensus sequence whatever itslength. Fromthese theoreticalpredictionsandthosefoundexperimentally,itclearly appearedthatthedistributionofarabinosidesandpolysaccharide substituentsalongthepolypeptidechainlead,contrarytowhatwas theoreticallypredicted,toacertainstiffnessofthebuildingblocks. Indeed,as suggestedby Goodrum etal. (2000),theamino acid sequenceofthe19-residuespeptideleadtoasymmetrical distri-butionofarabinosidesandpolysaccharidesubstituentsdirectedby thepalindrome-likearrangementoftheHypresiduesinthepeptide backbone.TheconformationalconsequenceisthattheAGP consen-susglycopeptideisslightlyrigidifiedbycontiguousHypblocks.This stiffnessfoundontheconsensusglycopeptidewouldbemuchmore pronouncedonthewholeAGPmacromoleculeandwouldeasily justifythelowerlevelofunorderedstructuresweexperimentally foundonintactandenzyme-cleavedAGPs.

It was however interesting to notice that the cleavage by proteasesofthepolypeptidebackboneinAGPdidnotaffectthe sec-ondarystructurecontent.Thisresultwouldbeinaccordancewith thepalindromicsequenceidentifiedinAGP(Goodrumetal.,2000), highlightingthefactthattheproteasecleavagewouldinducethe generationof verysimilar (poly)peptidefragments. Inaddition, thesimilaritiesfoundinthesecondarystructuresbeforeandafter hydrolyseswouldmeanthatglycomodulesattachedtothenative polypeptideandpolypeptidefragmentswouldbeverysimilarand wouldfavor againtherepetitiveand palindromicnatureofthe polypeptidesequencesfoundinintactandcleavedAGP.Indeed, Goodrumetal.(2000)gaveafinallengthforthepolypeptideof 380aminoacidswithabout20peptiderepeatsandoccasional par-tialrepeatstotakeintoaccountthehigherserinecontentinnative AGP.ThisrepetitionoftheconsensuspeptidetogetherwiththeHyp contiguityhypothesis(Shpak,Leykam,&Kieliszewski,1999)would leadtosimilar structuresfor theglycomodulesafterenzymatic hydrolyses.Theneworiginalresultsobtainedinthepresentstudy would mean that these similar (poly)peptide fragments would adoptthesameconformationastheuncleavedpolypeptide back-bone, illustrating a sort of self-ordering in thebuilding of the glycomodulesleadingtotheAGPmacromolecularassembly.This interestingresultwouldreinforcetheabsolutenecessityto per-forminanearfutureathoroughanalysisbymoderntandemmass spectrometryoftheexactpolypeptidesequenceofAGPafter degly-cosylationinordertoconfirmornottheexistenceofrepetitiveand palindromicsequencessuggestedintheliterature(Goodrumetal., 2000;Kieliszewski,2001; Kieliszewski&Shpak,2001).Tandem massspectrometrywouldalsobevery usefulfor the identifica-tionofthepresence ofoneorseveralpolypeptidesequencesin theAGPinordertoconfirmwhetherAGPisamolecular associ-ationofdifferentstructuralunitsassuggestedbyAl-Assafetal. (2009).

4. Discussion

Arabinogalactan(AG)andarabinogalactan-protein(AGP) frac-tionsfromA.senegalgumwerehydrolysedusingseveralproteases differingintheirmodeofactiontocleavepolypeptidebackbones. WeaimedtobettercharacterizethestructureofthecomplexAG andAGPfractionsbydeeplystudyingthesecondaryandtertiary structuresofthenativeandenzymaticallycleavedAGand AGP. EndoproteinaseGlu-C,pepsinandphosphataseacidwerethusused inacidicconditionswhilesubtilisinA,pronase,trypsin,papainand proteinaseKwereusedinalkalineconditionstocleavethetwo fractions.EnzymesdidnotaffectAGfractionstructurewhatever thepHconditionsused,highlightingtheinaccessibilityofthe pep-tidebackboneandtheremarkablestabilityofthisfractioninacidic andalkalineconditions.Thisresultwasinagreementwith previ-ouspublisheddata(Randalletal.,1989)andwiththethinoblate ellipsoidmodelwepreviouslyidentifiedfortheAGfractionwhere the43amino-acidresiduespeptidesequencewassupposed,based onspectroscopicmethods,tobetotallyburied(Renardetal.,2006; Sanchezetal.,2008).Theenzyme-treatedAGreportedhereinthis studywasfinallynotconclusiveintermsofnewstructural infor-mationandinparticularintheunderstandingoftheglycosylation arrangementsalongthepeptidebackboneduetotheabsenceof cleavageoftheshortpeptide.

Contrary to the remarkable stability of AG-peptide against hydrolysis,eveninalkalineconditions,numerousauthorsclearly demonstratedtheefficiencyofpronaseinthecomplete hydroly-sisoftheproteinintheAGPcomponentwithaten-folddecrease intheweightaveragemolecularmassofthemacromolecule( Al-Assafetal.,2009;Aokietal.,2007;Chikamaietal.,1996;Connolly etal.,1988;Duvalletetal.,1989;Elmananetal.,2008;Flindtetal., 2005;Mahendranetal.,2008;Osmanetal.,1993;Randalletal., 1989).Recently,itwasalsodemonstratedthatpapain-treated Aca-ciagumalsoshowedcompletedisappearanceoftheAGPfraction witha40%decreaseofitsoriginalweightaveragemolecularmass (Abdallaetal.,2012).However,areductioninMwwasalsonoticed

bytheauthorsinabsenceofenzymeduetothetemperatureapplied duringhydrolysiscastingsomedoubtsabouttherealefficiencyof papaintoreallydegradetheproteincomponentofAcaciagum.In addition,atoomuchhighsubstrate:enzymeratiocouldbeatthe originofthelackofenzymeefficiency.

Our present results, usinga substrate:enzymeratio of25 in alkalineconditions,indicatedthattheproteinmoietycouldbe effi-cientlydegradedwhatevertheenzymeusedincludingpapain.The lowestMwwasthereforereachedintreatedAGP-papainwithaMw

valueof1.68×105_gmol−1_{,aMw/Mnpolydispersityvalueof1.46}

andaradiusofgyrationvalueof11nm.TheMw/Mnpolydispersity

values were always slightly higher for enzyme-treated AGPs comparetocontrolAGPhighlightingthebroaderdistributionof AGPsfragmentsafterhydrolysisduepresumablytothedifferences in efficiency toward hydrolysis of the proteases used in these conditions(seeTable1)(Renardetal.,2012).Thedifferencesin enzymaticefficiencycouldalsoberelatedtolocalstericconstraints givingsomeenhanced accessibilityofcleavagesitestoenzymes and consequently increasingtheirefficiency. However,a single conformationin solutionclosetotheAGunit wasidentifiedin enzyme-treatedAGPsbyHPSEC-MALLS (see Fig.2).Theoverall structureoftheenzyme-treatedAGPswasfoundtobesurprisingly quitesimilarwhatevertheenzymeusedandclose,withhowever somesubtledifferences,toAGunit(seeFig.4).Atri-axial ellip-soidconformationwithsemi-axesvalues slightlydiffering from oneenzymetoanotherwasfoundinenzyme-treatedAGPs(see Table2).Moreinterestingly,thetwomainpreferentialdistances identified in thepair distancedistribution function(see Fig. 5) of control and enzyme-treated AGPs would claim in favor of rod-likeorelongatedparticlesoralternativelywouldindicatethe

presenceoftwoparticlesdifferingindimensions.These prelimi-narystructuraldatawouldthereforereinforcethehypothesisofthe molecularassociationsofstructuralunitsinAcaciagumthrough covalentlinkagesand/orhydrophobicassociationsandhydrogen bonding leading to the formation of the AGP macromolecular assembly.AnotherhypothesiswouldbethatAGPmacromolecular assembly, naturally present in exudate, would be intrinsically hydrolysedduringthematurationofnodulesbyexo/endogeneous enzymes,giving rise to molecular fractions withlower Mw, as

AG fractionfor instance. These structural units would thus be revealeddue totheirdissociationafterpolypeptidecleavage.As AGwaspreviouslyshowntobeunaffectedbyenzymetreatment inalkalineconditions,wemayeasilyconcludethatthehydrolysis onlyconcernedthepolypeptidesequenceofotherstructuralunits. AroughcalculationbasedonpreviousMwdeterminations(Renard

etal.,2006),andassumingthatAGPwouldbeamolecular associ-ationofAGandGPunits(Al-Assafetal.,2009),wouldgiveoneAG (Mw=2.86×105gmol−1)andonepolydisperseGPunits(average

Mw=1.59×106gmol−1) or one AG and three GP3 units (third

populationidentifiedby HPSEC-MALLS,Mw=2.95×105gmol−1)

to form AGP macromolecular assembly. However, the present studyrevealedthatstructuralunitsclosetoAGunitsbuthowever withsomestructuraldifferenceswouldbeinvolvedinthe forma-tionofAGPmacromolecularassembly.Inaddition,theverylarge polydispersityfoundinGPfraction(Renardetal.,2006)wouldcast somedoubtaboutitspossibleinvolvementintheformationofAGP macromolecularassemblyin particularbecauseseveral popula-tionswithMwhigherthanintactAGPwereidentified.Inaddition,

thedifferentresultsfoundinliteratureregardingthestructureof AGPfromAcaciagumwouldsuggestinanearfuturetogiveabetter termtodefinethismolecularfraction,AGPbeingoftenascribed toauniquemacromolecule(Al-Assafetal.,2009;Goodrumetal., 2000;Qi,Fong,&Lamport,1991;Renardetal.,2006,2012,2013).

Themolecularstructureofcontrolandenzyme-treatedAGPs surprisinglypredictedsimilarsecondarystructurescontentwith 11% ␣-helix, 25% ␤-sheet, 22% ␤-turn, 26% PPII helix and 16% unordered structure. These findings would indicate that the enzyme cleavage would induce the formation of similar (poly)peptidefragmentsidentifiedpreviouslyinAGPafter hydro-genefluoride(HF)-deglycosylationasbeingformedof19-residues repetitive and palindromic sequences (Goodrum et al., 2000). In addition, the similar conformations adopted by control and enzyme-cleaved AGPs would be in favor of a high regularity in the amino acidresidues sequence of the polypeptide back-bonein accordancewiththepalindromicnatureofthepeptide sequenceandtheoverallsymmetryoftheconsensusglycopeptide foundbyGoodrumetal.(2000).Indeed,authorsfoundthat palin-dromicsymmetryrigidifiedbycontiguoushydroxyprolinebuilding blocksmayimpartself-orderingpropertiesinAGP.Theseresults wouldquestionhoweveraboutthetypeofoligo-and polysaccha-ridebranchingoccurringontheAGPpolypeptidebackbone.This questionmotivatedsomesimplecalculationstodeterminethe the-oreticalmolecularweightofAGPtakingintoaccountthetypeof branchingonhydroxyproline(Hyp)residuesidentifiedbyQietal. (1991)and thoseidentified byGoodrum etal.(2000)extended toserine(Ser)andthreonine(Thr)residuesasdemonstratedby Mahendranetal.(2008).Thetheoreticalmolecularweights, tak-ingintoaccountourpreviousresultofthecompositionandthe totalnumberofaminoacidresiduespresentinAGP(Renardetal., 2006),wereof1.87×106_gmol−1_{onaQietal.(1991)basisforthe}

typeofbranchingandof1.61×106_gmol−1 _{onaGoodrumetal.}

(2000)basis forthetype of branching(Table 3).These calcula-tionssimplyhighlightedthefactthatHyp, ThrandSerresidues wouldbetheonlythreeaminoacidresiduesinvolvedinthe glyco-sylationmotifsthatdirectarabinosideoligomersandcarbohydrate blocksadditiontotheproteinbackboneofAGP.Theexistenceof

repetitiveandpalindromicsequencesintheproteinmoietyofAGP wouldthereforebeconfirmedthroughthesesimplecalculations asonly 4.7%differenceinthe theoreticalandexperimentalMw

wasfoundtakingintoconsiderationthetypeofbranching iden-tifiedbyQietal.(1991).Inaddition,thecalculationsperformed todeterminetheMw ofintactAGPconfirmedtheHyp

contigu-ity hypothesis which predicts arabinosides on contiguous Hyp residuesandarabinogalactan polysaccharidesonclustered non-contiguousHypresidues.Thesecalculations combinedwiththe structuralsimilaritiesfoundafterenzyme-treatedAGPwould rein-forcetheself-orderingarrangementsofthebuildingblocksalong thepolypeptidesequenceduetoacertainself-similararrangement duringassemblyoftheelementarybuildingblocks(i.e.the con-sensusglycopeptidefoundbyGoodrumetal.(2000),thatwecan definedastheelementarybuildingblock,canberepeatedntimes, onaself-similarbasis,toformwholeAGP,givingrisetoastructural self-similarityfortheAGPmacromolecularassemblywhateverthe lengthorobservationscale).Self-similarbehaviorwastherefore confirmedbySANS atalengthscalerangingfrom∼1to20nm, lengthscaleinaccordancewithsomeself-similarityarrangement ofelementarybuildingblocks.Beyondtheself-similarityor self-orderingarisingfromtheassociationofelementarybuildingblocks, whichwewillreferlatertobuilda3DmodelofAGP macromolec-ularassembly,we maywonderhow manyelementary building blocksconstituteAGPmacromolecularassembly.Toanswerthis question,weperformedcalculationstodeterminethenumberof shortandlongpolypeptidechainsinvolvedinAGP macromolecu-larassembly.Wemadetheassumptionthattwotypesofbuilding blocks would be present in AGPto perform thesecalculations. ThishypothesiscamefromtheidentificationbyMahendranetal. (2008) oftwo polypeptidechains bySDS-PAGE electrophoresis afterHF-deglycosylation ofAcaciagum,oneshortpeptidechain and one long polypeptide chain with 45 and 250 amino acid residues, respectively. The long polypeptide chain, withMw of

2.9–3_×104_gmol−1_{,correspondingto∼260–270residues,wasalso}

identifiedinAGPbyOsmanetal.(1993).Thecalculationswerethus performedtakingintoaccountthesetwopolypeptidechainsand thetypeofbranchingonHypresiduesdeterminedbyQietal.(1991) extendedtobranchingonSerandThrresiduesasdemonstratedby Mahendranetal.(2008).Finally,twoequationswithtwounknown parameters,xthenumberofshortpeptidechainsandythe num-beroflongpolypeptidechains,wereresolvedandgavex=6and y=8(Table4).AGPmacromolecularassemblywouldthusbemade oftheassociationof6glycomoduleswithMw=3.88×104gmol−1

and8glycomoduleswithMw=2.07×105gmol−1,each

glycomod-ule(i.e.buildingblock)beingmadeofarabinosideoligomersand polysaccharideblocksattachedtoHyp,SerandThrresiduesofthe shortandlong(poly)peptidechain.Itwasthereforeinterestingto noticethatthemolecularweightcalculatedforthebuildingblock 1wasclosetothosefoundexperimentallyafterchemical hydrol-ysisusingsodiumborohydride/sodiumhydroxidetreatmentsby Mahendranetal.(2008).Avalueof4.5×104_gmol−1 _wasfound

forthemolecularweightofthebuildingblockaftercompleteAGP degradation.Togofurtherinsightintothestructuralelucidation ofAGPmacromolecularassembly,wewantedtoknowhowmuch arabinosideoligomersandpolysaccharideblockswerepresenton each(poly)peptidechain.

Thenumberofbranches(arabinosideoligomersand carbohy-drateblocks)onAGPmacromolecularassemblywasdirectlygiven bythenumberofaminoacidresiduesinvolvedin(poly)saccharide substitutions(Table4).23and119Hyp,ThrandSeraminoacid residueswouldbeinvolved intheshortandlong(poly)peptide chain,respectively.Consideringthepercentagesofsubstitutions (arabinosideoligomersandcarbohydrateblocks)givenbyQietal. (1991)onHypresidues,and extended toSer andThrresidues, the number of branches for each building block and the total

D. Renard et al. / Carbohydrate Polymers 112 (2014) 648–661 657

Calculationsofarabinogalactan-protein(AGP)molecularweightMwandcomparisonwithexperimentalMwdeterminedbyHPSEC-MALLS.Calculationswerebasedontypeofbranchingonhydroxyproline(Hyp)residuesfrom Qietal.(1991)andGoodrumetal.(2000)extendedtobranchingonserine(Ser)andthreonine(Thr)residuesasdemonstratedbyMahendranetal.(2008).ThenumberofHyp,SerandThraminoacidresiduesweretakenfrom

Renardetal.(2006). Numberf residues involvedin oligo-and polysaccha-rides branching Typeof branchingon aminoacid residue(from Qietal., 1991) %Glycosylation %Glycosylation× numberof residues Aminoacid Mwinvolved inbranching Carbohydrate Mwinvolved in glycosylation TotalMw Typeof branchingon aminoacid residue(from Goodrum etal.,2000) %Glycosylation %Glycosylation× numberof residues Aminoacid Mwinvolved inbranching Carbohydrate Mwinvolved in glycosylation TotalMw NumberofHyp 766 Non-glycosylated 0.12 92 110 0 1.01E+04 Non-glycosylated 0.11 84 110 0 9.24E+03

Hyp-Ara3 0.64 490 110 440 2.70E+05 Hyp-Ara4 0.05 38 110 590 2.66E+04

Hyp-polysaccharide

0.24 184 110 4400 8.30E+05 Hyp-Ara3 0.27 207 110 440 1.14E+05

Hyp-Ara2 0.27 207 110 290 8.28E+04 Hyp-Ara 0.1 77 110 146 1.97E+04 Hyp-polysaccharide 0.2 153 110 4400 6.90E+05 NumberofSer 275 Non-glycosylated 0.12 33 110 0 3.63E+03 Non-glycosylated 0.11 30 110 0 3.30E+03

Ser-Ara3 0.64 176 110 440 9.68E+04 Ser-Ara4 0.05 14 110 590 9.80E+03

Ser-polysaccharide

0.24 66 110 4400 2.98E+05 Ser-Ara3 0.27 74 110 440 4.07E+04

Ser-Ara2 0.27 74 110 290 2.96E+04 Ser-Ara 0.1 28 110 146 7.17E+03 Ser-polysaccharide 0.2 55 110 4400 2.48E+05 NumberofThr 176 Non-glycosylated 0.12 21 110 0 2.31E+03 Non-glycosylated 0.11 19 110 0 2.09E+03

Thr-Ara3 0.64 113 110 440 6.22E+04 Thr-Ara4 0.05 9 110 590 6.30E+03

Thr-polysaccharide

0.24 42 110 4400 1.89E+05 Thr-Ara3 0.27 47 110 440 2.59E+04

Thr-Ara2 0.27 48 110 290 1.92E+04 Thr-Ara 0.1 18 110 146 4.61E+03 Thr-polysaccharide 0.2 35 110 4400 1.58E+05 Sumof aminoacid-oligo(poly)saccharides branching 1.75E+06 1.48E+06

SumofHyp,Ser,Thr residuesnotinvolvedin saccharidebranching 1.61E+04 1.46E+04 Sumofadditional non-glycosylated aminoacidresidues 1.14E+05 1.14E+05

Mw(calculated) 1.88E+06 1.61E+06

Mw(fromHPSEC-MALLS) 1.79E+06 1.79E+06

Table4

Calculationsofthenumberofbuildingblocks1and2involvedintheformationofarabinogalactan-protein(AGP)macromolecularassemblytakingintoaccountthetypeof branchingonhydroxyproline(Hyp)residuesdeterminedbyQietal.(1991)extendedtobranchingonserine(Ser)andthreonine(Thr)residuesasdemonstratedbyMahendran etal.(2008).Buildingblocks1and2werecomposedof45and250aminoacidresidues,respectively.ThenumberofHyp,SerandThraminoacidresiduesweretakenfrom

Renardetal.(2006).

Buildingblock1(withtheshortchainmadeof45aminoacidresidues) Numberofresidues involvedinoligo-and polysaccharides branching Typeof branchingon aminoacid residue(from Qietal.,1991) %Glycosylation %Glycosylation× numberof residues AminoacidMw involvedin branching Carbohydrate Mwinvolvedin glycosylation TotalMw NumberofHyp 15 Non-glycosylated 0.12 1 110 0 1.10E+02 Hyp-Ara3 0.64 10 110 440 5.50E+03 Hyp-polysaccharide 0.24 4 110 4400 1.80E+04 NumberofSer 6 Non-glycosylated 0.12 1 110 0 1.10E+02 Ser-Ara3 0.64 4 110 440 2.20E+03 Ser-polysaccharide 0.24 1 110 4400 4.51E+03 NumberofThr 4 Non-glycosylated 0.12 0 110 0 0.00E+00 Thr-Ara3 0.64 3 110 440 1.65E+03 Thr-polysaccharide 0.24 1 110 4400 4.51E+03 Sumofamino acid-oligo(poly)saccharides branching 3.64E+04

SumofHyp,Ser,Thrresiduesnot involvedinsaccharidebranching

2.20E+02 Sumofnon-glycosylatedadditional

aminoacidresidues

2.20E+03

Buildingblock1Mw(calculated) 3.88E4+O4

TotalAGPMw 1.88E+06

Totalnumberofresidues 2.25E+03

Numberofbuildingblocks1a ₆

Buildingblock2(withthelongchainmadeof250aminoacidresidues) Numberofresidues involvedinoligo-and polysaccharides branching Typeof branchingon aminoacid residue(from Qietal.,1991) %Glycosylation %Glycosylation× numberof residues AminoacidMw involvedin branching Carbohydrate Mwinvolvedin glycosylation TotalMw NumberofHyp 85 Non-glycosylated 0.12 11 110 0 1.21E+03 Hyp-Ara3 0.64 54 110 440 2.97E+04 Hyp-polysaccharide 0.24 20 110 4400 9.02E+04 NumberofSer 31 Non-glycosylated 0.12 4 110 0 4.40E+02 Ser-Ara3 0.64 20 110 440 1.10E+04 Ser-polysaccharide 0.24 7 110 4400 3.16E+04 NumberofThr 20 Non-glycosylated 0.12 2 110 0 2.20E+02 Thr-Ara3 0.64 13 110 440 7.15E+03 Thr-polysaccharide 0.24 5 110 4400 2.26E+04 Sumofamino acid-oligo(poly)saccharides branching 1.92E+05

SumofHyp,Ser,Thrresiduesnot involvedinsaccharidebranching

1.87E+03 Sumofnon-glycosylatedadditional

aminoacidresidues

1.25E+04

Buildingblock2Mw(calculated) 2.07E+O5

TotalAGPMw 1.88E+06

Totalnumberofresidues 2.25E+03

Numberofbuildingblocks2a ₈

a_TotalAGPM

w(calculated)=x(Mw(buildingblock1))+y(Mw(buildingblock2))andtotalnumberofresidues=x(45)+y(250)withx=numberofbuildingblocks1and

y=numberofbuildingblocks2.

numberofbuildingblocks1and2constitutingAGP macromolec-ularassembly,werefinallydetermined(seedetailsinTable5).Six buildingblocks1,withaMw=38880gmol−1,and8buildingblocks

2,withaMw=207000gmol−1,wouldcontributetotheassemblyof

thearabinogalactan-protein(AGP)fromAcaciagum.Themolecular associationsofthesebuildingblockswouldbedrivenbycovalent and/ornon-covalentinteractionsthroughtheinteractions occur-ringbetweenaminoacidresiduestolinkneighboringglycomodules

Fig.7.(A)Elementarybuildingblocks1and2usedtobuildarabinogalactan-protein (AGP)macromolecularassemblyfromAcaciaSenegalgum.Buildingblock2wasbuilt byalineararrangementoffivebuildingblocks1.(B)Three-dimensionalmodelof AGPmadeoftheassemblyof6buildingblocks1and8buildingblocks2bycovalent and/ornon-covalentinteractions.Theelementarybuildingblockswereconstructed takingintoconsiderationthequasi-palindromicnatureofpolypeptidebackbones

Table5

Calculationsofthenumberof(poly)saccharidesubstituents(arabinosideoligomers andcarbohydrateblocks)inbuildingblocks1and2,attachedtoHyp,SerandThr residuesoftheshortandlongpolypeptidechains,involvedintheformationof arabinogalactan-protein(AGP)macromolecularassembly.

ElementarybuildingblocksinvolvedinAGP macromolecularassembly

Number Mw

Buildingblock1(23aminoacidresiduesfor (poly)saccharidesubstitutions)a

-arabinosideoligomers 17

-carbohydrateblocks

6 TotalMw(shortpeptide

chain+(poly)saccharidesubstituents) (gmol−1₎

38880

Buildingblock2(119aminoacidresiduesfor (poly)saccharidesubstitutions)a

-arabinosideoligomers 87

-carbohydrateblocks

32 TotalMw(longpolypeptide

chain+(poly)saccharidesubstituents) (gmol−1₎

207000

Numberofbuildingblocks1inAGPb ₆

Numberofbuildingblocks2inAGPb ₈

CalculatedAGPMw(gmol−1) 1.88×106

ExperimentalAGPMwfromHPSEC-MALLS

(gmol−1)

1.79×106 a_{SeeTable4forthedetailsof(poly)saccharidesubstitutionsonHyp,ThrandSer}

aminoacidresidues.

b_{Number of building blocks 1 (or building blocks 2)=(M}

w arabinoside

oligomers/440)+(Mwcarbohydrateblocks/4400)withMwarabinosideoligomers

inbuildingblock1(or2)=0.73×MwsugarandMwcarbohydrateblocksinbuilding

block1(or2)=0.27×Mwsugar.Mwsugar(buildingblocks1and2)werecalculatedfrom

dataavailableinTable4withMwsugar(buildingblock1)=33880gmol−1andMwsugar

(buildingblock2)=179080gmol−1_{.Thevalues0.73and0.27correspondedtothe}

percentagesofglycosylationdeterminedbyQietal.(1991)butextendedto100%as sugarsmoietiesattachedtoHyp,SerandThrresidueswereonlyconsidered.

or through the numerous hydrogen bonds occurring between polypeptideand/or␤-1,3galactanbackbones.Athree-dimensional modelofAGPmacromolecularassemblywasbuiltbasedonthe repetitionofthepeptideconsensussequencefoundbyGoodrum et al. (2000) toget theshort and long polypeptidechains and theself-orderingprocessoccurringfromthesequasi-palindromic polypeptidesequencestoattachcarbohydratemoieties(Fig.7).The elementarybuildingblock2wasthereforebuiltfromtherepetition of∼5elementarybuildingblocks1linearlylinkedtomaintainthe palindromicnatureofthepolypeptidebackbone.Theself-similar approach wasthen usedtobuildthe3D modelof AGP macro-molecularassemblytakingintoconsideration theself-similarity ofthespatialorganizationofbuildingblocks2butalsothe spa-tialconfinementofbuildingblocks1and2tofillavolumeclose tothetri-axialellipsoidmorphologyweidentifiedbySANS. How-ever,thestatementthatagivenbiologicalstructureisself-similar shouldimplythattheiterativeprocessofitsconstructionhasareal biologicalmeaning,i.e.thatitsconstructioninnatureisachieved bymeansof asinglegenetic,enzymatic, orbiophysical mecha-nismsuccessivelyrepeated.Suchaniterativeprocesstoreproduce a self-similar objectwas successfully appliedtoglycogen, with atheoreticalfractaldimensionvalueof2(Melendez, Melendez-Hevia,&Canela,1999).Theexperimentaldfvalueof2.2probed

by SANS for AGP macromolecular assembly was in agreement withthedf values of2–2.5foundforhomogeneously branched

andasaconsequencetheoverallsymmetry,orself-ordering,ofcarbohydrate moi-etiesandarabinosideoligomersattachedtoit.Theself-similarityapproachwas thenusedtobuildtheAGPmacromolecularassemblyfromtheelementarybuilding blocks1and2.(C)Three-dimensionalmodelofpapain-treatedAGP.