Metabolomics and Cheminformatics Analysis of Antifungal Function of Plant Metabolites

15 Pages·2016·4.51 MB·English

H metabolites
OH
OH
Article
Metabolomics and Cheminformatics Analysis of
Antifungal Function of Plant Metabolites
MiroslavaCuperlovic-Culf*,NandhaKishoreRajagopalan,DanTulpanandMicheleC.Loewen
NationalResearchCouncilofCanada,1200MontrealRoad,Ottawa,ONK1A0R6,Canada;
[email protected](N.R.);[email protected](D.T.);
[email protected](M.C.L.)
* Correspondence:[email protected];Tel.:[phone]
AcademicEditor:PeterMeikle
Received:22August2016;Accepted:27September2016;Published:30September2016
Abstract: Fusariumheadblight(FHB),primarilycausedbyFusariumgraminearum,isadevastating
diseaseofwheat. PartialresistancetoFHBofseveralwheatcultivarsincludesspecificmetabolic
responsestoinoculation. Previouslypublishedstudieshavedeterminedmajormetabolicchanges
induced by pathogens in resistant and susceptible plants. Functionality of the majority of these
metabolites in resistance remains unknown. In this work we have made a compilation of all
metabolitesdeterminedasselectivelyaccumulatedfollowingFHBinoculationinresistantplants.
Characteristics,aswellaspossiblefunctionsandtargetsofthesemetabolites,areinvestigatedusing
cheminformaticsapproacheswithfocusonthelikelihoodofthesemetabolitesactingasdrug-like
molecules against fungal pathogens. Results of computational analyses of binding properties of
severalrepresentativemetabolitestohomologymodelsoffungalproteinsarepresented. Theoretical
analysishighlightsthepossibilityforstronginhibitoryactivityofseveralmetabolitesagainstsome
majorproteinsinFusariumgraminearum,suchascarbonicanhydrasesandcytochromeP450s. Activity
ofseveralofthesecompoundshasbeenexperimentallyconfirmedinfungalgrowthinhibitionassays.
Analysisofanti-fungalpropertiesofplantmetabolitescanleadtothedevelopmentofmoreresistant
wheatvarietieswhileshowingnovelapplicationofcheminformaticsapproachesintheanalysisof
plant/pathogeninteractions.
Keywords: Fusarium head blight; Fusarium graminearum; cheminformatics; metabolomics;
plantresistance;bioticstress;antifungal;carbonicanhydrase;cytochromeP450
1. Introduction
Fusarium graminearum is the primary pathogen causing Fusarium head blight (FHB) or scab,
a devastatingdisease primarilyaffecting wheatand barley crops inhumid andsemi-humid areas
worldwide. FHB causes major crop losses, as well as indirect losses due to grain contamination
causedbypotentmycotoxins,especiallydeoxyvalenol(DON)[1]. Strategiesfortheeradicationof
FHBincludefungicideapplications,culturalpractices,andthelonger-termdevelopmentofgenetic
resistance in the plant. Although there are several registered fungicides for application in wheat
and barley, none are very effective neither in the reduction of yield loss nor in the reduction of
mycotoxincontamination[1–3]. Cropswithgeneticresistancehavethehighestpotentialandarea
majorbreedingobjectiveworldwide. DNAmarkeranalysishasprovidedidentificationofquantitative
traitloci(QTL)thatprovidepartialresistancetosomebioticstressors,includingFHB[4]. Overone
hundredFHBresistance-associatedQTLshavebeenidentifiedinwheat,butthespecificfunctions
of these QTLs is still largely unknown [4,5]. Genetic control of resistance can induce biochemical
profilechangeleadingtotheresistanceresponse. Biochemicalresistanceisdirectlyassociatedwith
specificproteinsandmetabolites. Metabolomicsmethodshavebeenusedinanumberofstudiesto
Metabolites2016,6,31;doi:10.3390/metabo6040031 www.mdpi.com/journal/metabolites
Metabolites2016,6,31 2of15
determinemajormetabolicresponseinducedbyFHBinfectioninresistantandnon-resistantwheat
andbarley,showingsignificantconcentrationchangesinhormones,aswellasprimaryandsecondary
metabolites,asaresponsetoFHB.Analysisofmetabolicfactorsaffectingresistanceisprimarilyaimed
atdetectionofmajoreffectsofgenecontrolonmetabolismandpossiblenewfociforfurthergene
manipulation. Metabolomicsdatacanalsobeusedforthedeterminationofbiomarkersofresistance,
infection,andplantresponse. Finally,metabolomicsprovidesinformationabouttheresistance-related
metabolitesthatcaninducedevelopmentofbio-inspiredfungicidesorhelpdeterminefungalprotein
targetsforthedevelopmentofspecifically-targetedfungicides.
Anumberofsecondarymetabolites,includinghormones,phenolic,andpolyphenoliccompounds,
are significant in plant response to fungal infections. Hormones are metabolites that can easily
circulatethroughpartsof,orthroughthewholeorganismandthat,evenatlowconcentrations,signal
andcontrolresponses, growth, andthedevelopmentoforganisms[6]. Planthormones, including
auxins,cytokinins(CK),gibberellins(GA),abscisicacid(ABA),ethylene(ET),bassino-steroids(BR),
jasmonic acid (JA), and salicylic acid (SA), have major roles in plant defense against biotic and
abioticstressors[6,7]. SeveralfunctionsofSA,JA,andEThavebeendescribedingreatdetail[8,9].
Other hormones, including ABA, auxin, GA, CK, BR, and strigolactones, have been implicated as
important components of plant defense; however, their specific roles are not yet fully described.
SAplaysamajorroleinplantdefenseandisgenerallyinvolvedintheactivationofdefenseresponses
againstbiotrophicandhemi-biotrophicpathogenswhereSAlevelsincreaseinpathogenchallenged
plants. ExogeneousapplicationsofSAhavebeenshowntoinduceexpressionofpathogenesisrelated
genes (PR) [8,10–12]. JA and ET have stronger association with the defense against necrotrophic
pathogensandherbivorousinsects[13–15].AuxinandABAarealsoemergingashormonesinvolvedin
theregulationoftheresponsetobothbiotrophicandnecrotropicpathogens,althoughthemechanisms
oftheiractionarestillnotclear[7,16–19]. GAhormoneisproducedbyplants,fungi,aswellasbacteria,
andappearstoalsohaveasignificantroleinplantdiseaseprogression,butthespecificfunctionhas
notbeendescribed[20].
Therolesofphenoliccompounds(includingphytoalexins)infungalresponsesofplantshave
beenreviewedbyLattanzioetal.[21]. Phenoliccompoundsrefertoallsecondarynaturalmetabolites
arising from the shikimate-phenylpropanoids-flavonoids pathways. Antifungal phenolics, which
includetanninsandproanthocyanidins,canbeeitherpreformed(phytoanticipins)andredistributed
following infections, or can be synthesized following interaction between the host and fungal
parasite (phytoalexins). Over the last several years a number of metabolomics studies have been
performed in order to determine plant metabolites associated with fungal infection, including
FHB. Mass spectrometry was used for study of metabolic changes following FHB infection in
resistant barley genotypes [22,23] as well as resistant wheat cultivars [5,24,25]. Nuclear magnetic
resonance-basedmetaboliteprofilingwasutilizedforscreeningofpassiveresistanceinwheatagainst
FHB[26]. Finally,severalvolatileorganiccompoundsproducedbychickpeasshowstronganti-FHB
activity[27]. Theeffectofasmallnumberofthesemetaboliteshasbeentestedinvitro. Phenoloic
acids,includingferulicacid,arethemostabundantinwheatbran. Followinginfection,phenolicacids
areover-concentratedinresistantwheatandtheyhavebeendemonstratedasefficientinhibitorsof
mycotoxinproductioninseveralstrainsofFusarium[28,29]. Antifungalandantibacterialactivityof
tannicacidhasbeenknownforalongtime[30–32],withactivityagainstFusariumgraminearumshown
recently[33].
Thegoaloftheworkpresentedherewastodeterminewhetherplant-producedmetabolitescan
act as drug-like agents of the plant’s biotic response, and what are their possible protein targets.
Cheminformatics analysis was used to identify plant metabolites with the highest potential to
functionasfungalgrowthinhibitorsand,subsequently,theeffectofseveralselectedmetaboliteswas
testedexperimentally. Wepresenthereacompilationofallpreviouslydeterminedresistance-related
metabolites,followedbycheminformaticsanalysisofphysicochemicalpropertiesandtherelationships
betweenthesecompounds. Possibleproteintargetsofsomeofthesecompoundshavebeenpredicted
Metabolites2016,6,31 3of15
usingpubliclyavailabledataforthese,andrelated,molecules. Theinteractionpotentialbetweenplant
metabolitesandproteinsinfungusisexploredcomputationallyusinghighthroughputsmallmolecule
docking. Subsequently,selectedmetaboliteshavebeentestedexperimentallyshowingactivityagainst
Fusariumgraminearuminfungalgrowthinhibitionassays.
2. ResultsandDiscussion
2.1. SelectionofMetabolitesandCheminformaticsAnalysisofTheirMolecularandDrug-LikeProperties
Metabolic changes resulting from FHB infection in resistant or less susceptible plants have
been explored in wheat [5,24,25,32–34] and barley [23,35]. Anti-fusarium metabolites have also
been explored in chickpeas, and their effect was tested for protection of wheat from FHB [27].
Additionally, investigations of metabolites commonly involved in biotic responses [21,36], as well
as plant hormones with possible roles in resistance [7,18,37] have also been performed. Although
thereissomeoverlapinmetabolitesthataccumulateasaresponsetobioticstressacrossthesedistinct
experimentsandacrossdifferentspecies, therearealsomanyuniquemetabolitesidentifiedinthe
differentreports. Discrepanciesinobservedmetabolicprofilescanbearesultofdifferentbehaviors
ofplantsovertime,indistinctconditions,orindifferenttissues. Metaboliccoveragealsodepends
on experimental procedures. Finally, continual improvements in sensitivity of experimental and
analytical methods can introduce new metabolites to the measurements. In this work we have
made a compilation of all metabolites determined in previously published studies dealing with
FHB infection in grains. A complete list of these resistance related (RR) metabolites is provided
in Supplementary Table S1. In total, 478 RR metabolites have been included in our database.
Of478RRmetabolites474areregisteredwithinPubChem[38]and,thus,haveaccessiblestructural
information. SupplementaryTableS1includesPubChem[38]CIDsforthese474metabolites,aswell
as compounds’ names provided in original references or in PubChem [38] for all RR metabolites.
FormetabolitesthathavenotbeenregisteredinPubChem[38]SupplementaryTableS1showsIDs
foreitherKNApSAcK[35],HMDB[39],andChEBI[40]databases. Thesemoleculesbelongtomany
differentgroups,includingphenols,polyphenols,flavonoids,etc. Moleculescanbeclassifiedbased
ontheirchemicalcharacteristics,functionalgroups,size,biologicalpathwaysoforigin,etc.,andfor
manyofthemetabolitespresentedthesetypesofclassificationareavailableintheoriginalliterature.
Figure1AshowsthecalculatedvaluesofseveralsignificantmolecularcharacteristicsforRRmetabolites
comparedtothoseforallknownplantmetabolites(fromPubChem[38]andGolmdatabases[41]),
suchasthemolecularweight(MW),thenumberofbothhydrogenacceptorsanddonors(HBAand
HDA),andthenumberofheavyatoms(HA).
InorderforRRmetabolitestoactasplant-producedantifungalagentstheymusthavedrug-like
properties that can be simply represented through Lipinski “Rule-of-five” characteristics [42,43].
Lipinskirule-of-fiveisastandardmethodindrugdiscoveryanditsutilityinantifungaldevelopment
hasbeenrecentlyshown[43]. AcomparisonofLipinskirulesbetweenRRmetabolites,allknownplant
metabolites,andFDAapproveddrugs(Figure1B–D)allowsaquickinvestigationofthepossibility
for plant selection of drug-like characteristics in RR metabolites. A number of other molecular
characteristicscalculatedforthisset(intotal251propertiescalculatedusingRCDK)includingphysical,
chemical, and electronic properties again show similar ranges in the two groups of metabolites
(dataavailablefromtheauthors). RRmetabolitesspanawiderangeofmoleculesfromthehighest
molecularweightmetabolites—tannicacid(MW=1701.198;includedasageneralresistancerelated
metabolite[21])andPhyllanthusminB(MW=926.6506;producedbyresistantwheat)—tothelowest
molecular weight molecule, ethylene (general plant hormone). The size and number of hydrogen
bonddonorsandacceptorsforRRmetabolitesareforthelargemajorityofRRmetaboliteswithinthe
Lipinskirule-of-fivefordrug-likemolecules(Figure1A).
Metabolites2016,6,31 4of15
Metabolites 2016, 6, 31 4 of 15
Figure 1. (A) Range of molecular sizes and structures in resistance-related (RR) metabolites,
Figure1.(A)Rangeofmolecularsizesandstructuresinresistance-related(RR)metabolites,including
including the molecular weight (MW), the number of bonds (nBonds), the number of CH3 groups
the molecular weight (MW), the number of bonds (nBonds), the number of CH3 groups (nCH3),
(nCH3), and the number of rotatable bonds (nRotBonds); (B–D) show values for Lipinski
and the number of rotatable bonds (nRotBonds); (B–D) show values for Lipinski “rule-of-five”
“rule-of-five” characteristics [42] of molecules in RR, plant metabolites (Metab), and FDA-approved
characteristics[42]ofmoleculesinRR,plantmetabolites(Metab),andFDA-approveddrug(FDA)sets;
drug (FDA) sets; (B) shows values of the logarithm of the partition coefficient for n-octanol/water
(B)showsvaluesofthelogarithmofthepartitioncoefficientforn-octanol/waterLogP(XLogP)relative
LogP (XLogP) relative to MW; (C)topological polar surface area (TPSA) of molecules relative to MW
toMW;(C)topologicalpolarsurfacearea(TPSA)ofmoleculesrelativetoMWand(D)ZagrebIndex
and (D) Zagreb Index values (ZgIndex) relative to MW. Stars in (B–D) show the average values for
values(ZgIndex)relativetoMW.Starsin(B–D)showtheaveragevaluesforallmoleculesinthegroup
all molecules in the group with black indicating FDA, red indicating RR, and blue indicating Metab.
withblackindicatingFDA,redindicatingRR,andblueindicatingMetab.Thelargersimilaritybetween
RTRhea nladrgFeDrA simmoillaerciutyle sbiestwapepeanr eRnRt inanadll FcaDsAes smuoglgeecsutliensg itsh aatpRpRarmenett aibno aliltle scahsaevse siungcrgeeassteindgd rthuagt- liRkRe
qmueatlaitbieosli,tpesa rhtiacvuel airnlcyreinasteedrm dsruogf-pliakses iqvueamlietimesb, rpaanretitcrualnasrplyo ritnr teelramtivse otfo paavsesrivage empelmanbtrmaneeta tbroanlistep.ort
relative to average plant metabolite.
Thedrug-likecharacteristicsofRRmetabolitesareexploredincomparisontoallknownplant
The drug-like characteristics of RR metabolites are explored in comparison to all known plant
metabolitesintheGolmdatabase[41]andallFDAapproveddrugsincludedintheZINCdatabase[44].
metabolites in the Golm database [41] and all FDA approved drugs included in the ZINC database
ForthisanalysisweselectedfromPubChem[38]allRRmetabolitesthathaveavailable3Dstructures,
[44]. For this analysis we selected from PubChem [38] all RR metabolites that have available 3D
aswellasanyplantmetabolitesprovidedbytheGolmdatabase[41]with3Dstructures.FDA-approved
structures, as well as any plant metabolites provided by the Golm database [41] with 3D structures.
drugs with 3D structures from ZINC drug database were also selected [44]. This yielded 416 RR
FDA-approved drugs with 3D structures from ZINC drug database were also selected [44]. This
metabolites,1171plantmetabolites,and1985FDA-drugs. Forthesemetaboliteswehavecalculated
yielded 416 RR metabolites, 1171 plant metabolites, and 1985 FDA-drugs. For these metabolites we
severalphysicochemicalproperties,includingoctanol/waterpartitioncoefficient(logP),molecular
have calculated several physicochemical properties, including octanol/water partition coefficient
weight (MW), topological polar surface area (TPSA), and Zagreb Index using the RCDK package
(logP), molecular weight (MW), topological polar surface area (TPSA), and Zagreb Index using the
inR[45](RCDKwasdevelopedbyR.Guha),andcomparisonsbetweentheRRmetabolites,plant
RCDK package in R [45] (RCDK was developed by R. Guha), and comparisons between the RR
metabolites(Metab),andFDAdrugsareshowninFigure1B–D.Plotsshow(asfilledyellowstars)
metabolites, plant metabolites (Metab), and FDA drugs are shown in Figure 1B–D. Plots show (as
averagevaluesforeachgroupofmetabolites. Althoughcharacteristicsofmoleculesinallthreegroups
filled yellow stars) average values for each group of metabolites. Although characteristics of
arecomparable,averageMW:logP,MW:TPSA,andMW:ZagrebIndexforRRandFDAmoleculesare
molecules in all three groups are comparable, average MW:logP, MW:TPSA, and MW:Zagreb Index
morecloselyrelated,suggestingthatRRmetabolitesgenerallyhavemoredrug-likecharacteristics
for RR and FDA molecules are more closely related, suggesting that RR metabolites generally have
thangeneralplantmetabolites.
more drug-like characteristics than general plant metabolites.
The majority of RR metabolites have a larger average partition-coefficient (XlogP) value than
what is seen in the overall plant metabolite set indicating their higher lipophilicity. XlogP is an
atom-additive method for calculating the octanol/water partition coefficient (logP). It obtains the
Metabolites2016,6,31 5of15
ThemajorityofRRmetaboliteshavealargeraveragepartition-coefficient(XlogP)valuethanwhat
isseenintheoverallplantmetabolitesetindicatingtheirhigherlipophilicity.XlogPisanatom-additive
methodforcalculatingtheoctanol/waterpartitioncoefficient(logP).ItobtainsthelogPvaluefora
compoundbysummingthecontributionsfromcomponentatomsandcorrectionfactors. Lipinski’s
rule-of-fiveallowedrangeforXlogPis−0.4to5andthemajorityofRRmetabolitesfitwithinthese
limits(Figure1B).Thetopologicalpolarsurfacearea(TPSA)ofmoleculesdefinesthesumofsurfacesof
polaratomsinamolecule(calculatedusingthemethoddevelopedby[46]). Inthecurrentlyusedform,
moleculeswithTPSAlargerthan140Å2arebelievedtohavealowcapacityforunaidedpenetration
throughcellmembranes,whilethosewithTPSAlessthan60Å2 areeasilyabsorbedbythecellular
membrane[47]. MoleculeswithTPSAover60Å2arebelievedtobefavoredandfurtherregulatedby
moleculartransporters,suchasABCtransporters[48]. TheaverageTSPAofRRmetabolitesislarger
thanforoverallplantmetabolitesandhighlycomparabletotheaverageforFDA-approveddrugs
(Figure1C).ItshouldbekeptinmindthatABCtransportersarecrucialpathogen-relatedproteins
highly overexpressed and involved in FHBresistance[49], possibly suggestingtheir role in active
transport of RR metabolites. Figure 1D compares values for the Zagreb Index between RR, FDA,
andplantmetabolites. TheZagrebIndex,isatopologicalmeasureofmolecularbranching,calculated
asthesumofthesquaredvertexvalences,i.e.,thenumberofconnectionstoheavyatomsregardlessof
theirbondorder[50]. Overall,theZagrebIndexvalues(ZgIndex)aresignificantlyhigherforRRthan
plantmetabolites,andhighlysimilarbetweenRRandFDAmolecules,althoughvaluesforallthree
groupsarestillwithinthepreviouslyproposed,verylarge,optimalrangeof22–452[51].
2.2. MetaboliteActivity
Anumberofbiologicalassayshavealreadybeenperformedformanyofthemetaboliteswithin
theRRgroup. Proteintargetsarealsoknownformanyoftheseandthisdataisavailablethrough
thePubChemdatabase[38]. Thisinformationcanguideusinproposingpossiblefunctionsoftested,
aswellasrelated,moleculesinthebioticresponseofaplant. KnownactivitiesforRRmetabolites
areshowninFigure2. Accordingtothepubliclyavailabletests,serotoninisthemostpromiscuous
compound,showingbindingwiththelargestnumberoftestedproteins. Thelargestnumberoftested
compoundstargetcarbonicanhydraseproteins.
The RR metabolites show activity against a number of proteins and the complete list can be
obtainedfromtheauthors. Anumberofmetaboliteshavemorethanoneknownproteintargetandfor
severalproteintargetsmorethanonemetaboliteshowsactivity(Figure2).
AnumberofactivemetaboliteslistedinFigure2havebeenpreviouslyshownasantifungals.
General antifungal and antibacterial activity of tannic acid, for example, has been known for
a long time [30–32] and activity against Fusarium graminearum was recently experimentally
shown[34]. Ouranalysispresentsseveralpossibleproteintargetsfortannicacid,includingcarbonic
anhydrasesandabromodomainadjacenttothezincfingerdomain2B,bothhavinghomologinthe
Fusariumgraminearumsequence. Manyothertargetsoftannicacidhavebeendefinedintheliterature.
WhiletheactivityoftannicacidagainstFusariumgaminearumhasalreadybeenreported,itservesasa
goodpositivecontrol. Atthesametime,tannicacidisalargephenoliccompoundthathasstructural
similaritytomanyother,simplerphenolicacids,detectedingrainsinfectedwithFHB.Examplesof
theseincludeferulicacid,p-coumaricacid,salicylicacid,gallicacid,caffeicacid,etc.Thesesignificantly
smallerproductsofthephenylpropanoidbiosyntheticpathwaypossiblyhavesimilarproteinbinding
targetsastannicacid.
Metabolites2016,6,31 6of15
Metabolites 2016, 6, 31 6 of 15

Figure 2. Known protein targets of RR metabolites. Shown are high-activity relationships where (A)
Figure 2. Known protein targets of RR metabolites. Shown are high-activity relationships where
shows metabolites with known activity against known protein targets; and (B) protein targets listed
(A)showsmetaboliteswithknownactivityagainstknownproteintargets; and(B)proteintargets
with more than four active metabolites determined from the subset of metabolites measured within
listedwithmorethanfouractivemetabolitesdeterminedfromthesubsetofmetabolitesmeasured
the RR set. Protein names correspond to the UniProt database [52] names and are listed in the figure
withintheRRset.ProteinnamescorrespondtotheUniProtdatabase[52]namesandarelistedinthe
legend.
figurelegend.
Other highly active, tested metabolites include luteolin and serotonin, both structurally related to
a Ontuhmerbherig holyf aocttihveer, teRsRte dmmeettaabbooliltietes.s inAclltuhdoeugluht eothlienira nadcstievriotiteosn inin, bobtihoasstrsuayctsu rahlalvyer elbaeteedn t o
a neuxtmenbseivreolyf sottuhdeiredR, Rthmeire teaffbeoctl itaegsa.inAst ltFhuosaurgiuhmt ghreaimrianceatirvuimti ehsasi,n tob itohaes bsaesyts ohf aovuer kbneoewnleexdtgeen, sniovte ly
stubdeieend e,xtphleoirrede.f fLeuctteoalgina,i nass twFeulls aarsi usemvegrarla motihneera rfluamvonheass ,wtioth tdhieffebreesntt hoyfdoruorxyklantoiown lpedatgteer,nns,o htabvee en
expbleoerne dp.reLvuiotueosllyin ,idaesnwtifeieldl asass eRvRe ramlooltehceurlefls awvoitnhe ssowmiteh pdrioffteeirne nttahrgyedtsr oaxlyrelaatdiyo nidpeantttiefriends ,fhoar v e
beeanpigperneivni oanudsl ykaeidmepnfteirfioel d(FiagsurRe R2).m Tohleesceu, laess wwelilt has ssoemveeralp orothteerin luttaeorglient-srelaaltreeda dmyoleidcuenletsi,fi headvef or
apibgeeenni onbasenrdvekda ienm inpffeecrteodl (wFhigeuatr e[5,22)3.]. TLuhteesoel,ina iss wa heilglhalsy speovteenrat lanottihoxeirdlauntte aonlidn r-raedliactael dscmavoelnegceurl, es,
havsiembileaerlny obtos erovtheder inflianvfoencoteidds.w hLiekaetw[i5s,e2 3]t.o Louttheeorl inmiestaabohliigtehsl,y tphoet enacttiavnittyio xoidf alnutteaonlidn rahdaisc a l
scabveeenng epr,rismimariillayr lytestotedo thaegraiflnastv onmoaimdsm.alLiaink ewpirsoetetions.o thOenrcem eatagbaionl,i tems,atnhye aocft ivtihtey opfrolutetienosl i n
hasthbate elnuteporliimn airsi lyacttievset eadgaaignasti nhstavme acmlomsea lpiarnotepinro tseeiqnuse.ncOe nhcoemaoglaoigns, imn aFnuysaorifumth egrparmoitneeianrsumth. a t
It is also interesting that glycosylated luteolin gets actively transported by ABC-type transporters [53].
luteolin is active against have close protein sequence homologs in Fusarium graminearum. It is
Luteolin is structurally related to a number of other flavones and flavone derivatives
also interesting that glycosylated luteolin gets actively transported by ABC-type transporters [53].
within our RR metabolites set (e.g., isovitexin-7-O-xyloside, isovitexin 2″-O-(6′′′-feruloyl)glucoside,
Luteolin is structurally related to a number of other flavones and flavone derivatives within
ouirsoRviRtexmine-7t-aOb-oglliutecsosysle-t2″O(e-.rgh.a,minsoovsiidteex, iann-d7 -Oka-exmylpofseirdoel-,3-gisluocvoitseidxein-7-r2h”a-mOn-(o6s(cid:48)i(cid:48)d(cid:48)-ef,e rtuo lonyaml)gel ujucsot said e,
few).
isovitexin-7-O-glucosyl-2”O-rhamnoside, andkaempferol-3-glucoside-7-rhamnoside, tonamejust
Serotonin (5-hydroxytryptamine) is an extensively studied neurotransmitter in mammals, as
afew).
well as a widely-distributed metabolite in plants. Serotonin is highly structurally similar to auxin
Serotonin(5-hydroxytryptamine)isanextensivelystudiedneurotransmitterinmammals,aswell
(indol-3-acetic acid), one of the major plant hormones. It has been hypothesized that serotonin acts
as a widely-distributed metabolite in plants. Serotonin is highly structurally similar to auxin
as an auxin function inhibitor, thereby regulating root development [54]. In resistant wheat infected
(indol-3-aceticacid),oneofthemajorplanthormones. Ithasbeenhypothesizedthatserotoninacts
with Fusarium graminearum a concentration increase has been observed in hydroxycinnamic acid
asanauxinfunctioninhibitor,therebyregulatingrootdevelopment[54]. Inresistantwheatinfected
amide-conjugated putrescine, tyramine, agmantine, and serotonin acting as pytoalexins [5].
witSherFoutosnariniu mandgr aimts indeearriuvmativaecso np-cceonutmraatrioylnseirnoctroenaisne ahnads bfeereunlooyblsseerrovtoendinin wheyrde roaxccyucminunlaamtedic ianc id
amBidipeo-lcaorins jourgyaztae idnfpeuctterdes rciicnee [,5t5y]r awmhienree ,saegromtoannitnin ter,eaatnmdesnetr osutopnpirnesascetsi nthgea gsrpoywtotha loefx finusng[5a]l. hSyeprohtaoen. in
anditsderivativesp-coumarylserotoninandferuloylserotoninwereaccumulatedinBipolarisoryza
Metabolites2016,6,31 7of15
infectedrice[55]whereserotonintreatmentsuppressesthegrowthoffungalhyphae. Severalother
Metabolites 2016, 6, 31 7 of 15
metaboliteswithintheRRgroupalsocontainanindoleringandarehighlysimilartoserotonin.
Several other metabolites within the RR group also contain an indole ring and are highly similar to
2.3. ProteinTargets
serotonin.
Asprevioustestingofbiologicalactivityofcompoundswasprimarilyperformedonmammalian
2.3. Protein Targets
proteins,wehaveutilizedsequencesimilaritystudiesandhomologymodellingtoestablishprotein
As previous testing of biological activity of compounds was primarily performed on
functionsinwheatandFusariumgraminearumsystems. Outof327proteintargetsobtainedaspossible
mammalian proteins, we have utilized sequence similarity studies and homology modelling to
action sites for the tested RR metabolites, 211 are Homo sapiens proteins. In order to determine
establish protein functions in wheat and Fusarium graminearum systems. Out of 327 protein targets
relevanceofthisfindingtointeractionsbetweenplantsandFusariumgraminearumwehavemappedthe
obtained as possible action sites for the tested RR metabolites, 211 are Homo sapiens proteins. In order
completteo hduetmeramninper roetleevoamnceea ogfa tihniss tfipnrdointegi nto sienqteuraecnticoenss abveatwileaebnl epflaonrtsF uansadr Fiuumsargiurmam girnameairnueamru(ms twraei nPH-1;
data obhtaaivnee mdafprpoemd tthhee cBormopaldeteI nhsutimtuante )p.roStetoamtiset iacgaalininstf oprrmoteaitni osneqfuoerncceos mavpalielatebleh ufomr aFnusatoriucmo mplete
Fusariumgragmrianmeairnuema r(usmtraipnr oPtHeo-1m; deamta aopbptainingedi sfrsohmo wthne inBrFoaigdu Irnest3it.utOe)u. tStoaftisatlilca8l 9i,n0f3o3rmhautimona nfopr roteins
complete human to complete Fusarium graminearum proteome mapping is shown in Figure 3. Out of
obtained from UniProt database [52], 36% could be mapped to the 13,321 Fusarium graminearum
all 89,033 human proteins obtained from UniProt database [52], 36% could be mapped to the 13,321
proteinsobtainedfromtheBroadInstitutedatabase[56].
Fusarium graminearum proteins obtained from the Broad Institute database [56].

Figure 3. Statistical information regarding the mapping of human proteins to the Fusarium
Figure3.StatisticalinformationregardingthemappingofhumanproteinstotheFusariumgraminearum
graminearum proteome. Sequence similarity was determined using in-house methodology.
proteome.Sequencesimilaritywasdeterminedusingin-housemethodology.
For the subset of 211 human proteins shown as targets of tested plant metabolites, 138 have
Forsigthneificsaunbt sehtomofol2o1g1s hinu mthea nFupsraorituemin sgrsahmoinweanrumas steaqrugeentcse.o fThteesretefodrep, lwanitthimn etthaeb osulibtseest, o1f3 8 have
metabolite-targeted proteins, 65% have homologs in Fusarium graminearum.
significant homologs in the Fusarium graminearum sequence. Therefore, within the subset of
metabolite-targetedproteins,65%havehomologsinFusariumgraminearum.
Homology Modelling and Docking Analysis—Activity of Metabolites against Fungal Targets
The majority of proteins listed as targets in Figure 2 have highly similar Fusarium graminearum
HomologyModellingandDockingAnalysis—ActivityofMetabolitesagainstFungalTargets
protein homologs based on the sequence comparison performed in this work. For example, CA
TheFumsaarijuomri tgyraomfinperaorutmei nsesqluisetnecde aFGsStaGr_g0e4t6s03in hFasi gtuher ea2lphhaa vCeAh siguphelyrfasmimilyil adromFuasina raiundm cglorasemsti nearum
proteinsihmoimlaroitlyo gtos hbuamsaend CoAn (BthLAeSsTe qscuoeren: c7e ×c 1o0m−20p). aCraisrobonnipc earnfhoyrdmraesde (iCnA)t hisi sa wknoorwkn. taFrgoert oefx ample,
tannic acid [57], serotonin, ferulic acid [58], as well as number of other RR metabolites either through
CAFusariumgraminearumsequenceFGSG_04603hasthealphaCAsuperfamilydomainandclosest
specific or non-specific binding and inhibition. At the same time CA has been indicated as an
similaritytohumanCA(BLASTscore:7×10−20).Carbonicanhydrase(CA)isaknowntargetoftannic
interesting target for antifungal agents (recently reviewed in [59]). Although several indicated
acid[57],serotonin,ferulicacid[58],aswellasnumberofotherRRmetaboliteseitherthroughspecific
metabolites have experimentally shown binding to human CA, binding to fusarium CA was not
or non-stepsetecdi fifocr bainnyd oinf tghea nRdR minehtaibboitliitoens.. InA otrtdheer tsoa emxpelotriem teheC pAosshibalse bbeinedninign dofi cRaRt emdetaasboalniteisn ttoe resting
targetfofursaanrituimfu nCgAa wlae gheanvets p(errefocremnetldy hriegvh-itehwroeudghinpu[t5 c9o]m).pAultathtioonuagl hdoscekvinegra alnianlydsiicsa. tedmetaboliteshave
experimentaFlolry dsohcokiwngn Fbuisnadriuinmg gtroamhinuemaraumn CCAA ,wbaisn mdiondgeltleodf uussianrgi uthme hCoAmowloagsy nmootdteelslitnegd mfoerthaondy ofthe
provided by Swiss-Model [60–62], in comparison to proteins with highest sequence similarity with
RRmetabolites. InordertoexplorethepossiblebindingofRRmetabolitestofusariumCAwehave
available X-ray 3D structures. Fusarium graminearum sequence FGSG_04603 was used as it was the
performedhigh-throughputcomputationaldockinganalysis.
closest homolog to several isozymes of human carbonic anhydrases of known structure.
FordocAk ninugmbFeurs oafr Xiu-rmayg crraymstainl estarruucmturCesA arwe aavsaimlabolde efollre CdAu psriontgeinths,e inhcloumdionglo rgepyremseondtaetlilviensg ofm ethod
provideddifbfeyreSnwt cilsasss-Mes oasd wele[ll6 a0s– 6d2is]t,ininct cisoomzypmaersi sforonmt ohupmroante cienlsls.w Sietvherhail gahvaeislatbsleeq curyesntacle stsriumctiularersi tywith
availabliencXlu-drae yin3hDibitsotrryu mctoulreecus.leFs.u Dsoacrkiuinmg agnraalmysiinse faorru amll RsReq mueetanbcoeliFteGs wSGas_ p0e4r6fo0r3mweda fsoru hsuedmaans CitAw asthe
for comparison and Fusarium graminearum CA. Human CA I in complex with the inhibitor
closesthomologtoseveralisozymesofhumancarbonicanhydrasesofknownstructure.
topiramate (pdb: 3LXE) was selected for the analysis of human protein. This is the structure of one of
AnumberofX-raycrystalstructuresareavailableforCAproteins,includingrepresentativesof
the known target proteins of serotonin and contains an inhibitor, allowing the determination of the
differentclassesaswellasdistinctisozymesfromhumancells. Severalavailablecrystalstructures
include inhibitorymolecules. DockinganalysisforallRRmetaboliteswasperformedforhumanCA
forcomparisonandFusariumgraminearumCA.HumanCAIincomplexwiththeinhibitortopiramate
(pdb: 3LXE) was selected for the analysis of human protein. This is the structure of one of the
Metabolites 2016, 6, 31 8 of 15
grid zone for docking analysis. Fungal CA homolog FGST_04603 was modeled using crystal
structure pdb: 3Q31 of the fungal alpha carbonic anhydrase from Aspergillus oryzae. Carbonic
anhydrase from Aspergillus oryzae showed 39.32% sequence identity with fusarium graminearum CA,
studied here.
Figure 4A shows Ramachandran plots of the 3D structure of human carbonic anhydrase I
(3LXE) and the homology model of protein FGSG_04603 is modeled using the 3Q31 crystal structure.
Metabolites2016,6,31 8of15
Figure 4B shows a structural comparison between human carbonic anhydrase 1 (3LXE) and the
model of Fusarium graminearum carbonic anhydrase FGSG_04603 in space filling and skeletal format.
knowntAa rsgkeetleptarlo tfoerimnsato fcosmerpoatroens ianctaivned scitoens toafi nthsea ncairnbohnibici taonrh,yadllroawse inpgrottehiensd feotlelorwmiinnga tsitorunctoufret hegrid
alignment. Additionally, the inhibitor (green), in order to mark the active site of the two proteins, is
zone for docking analysis. Fungal CA homolog FGST_04603 was modeled using crystal structure
shown.
pdb:3Q31ofthefungalalphacarbonicanhydrasefromAspergillusoryzae. Carbonicanhydrasefrom
Structure quality has been assessed using Ramachandran plots for the model (Figure 4A). A
Aspergillcuosmoprayrzisaoens hoof wReadm3ac9h.3an2d%rasne qpuloetns cefoird ethnet it3yLXwEi thstrFuucstuarreiu mandg rathmei nmeaordueml sChAow, stau dsiimedilahr ere.
Figduirsetri4bAutisohno owf sreRsiadmueasc hina nstdruractnurpall ogtrsooufpsth oef 3thDe sβt-rsuhceteut,r aeso wfehlul mas atnhec alerfbt-o annidc aringhhyt-dharnadseedI (3LXE)
andtheαh-ohmelioxl.o Ag ylamrgoe dmealjoorfitpyr (ootveeinr 9F0G%S) Gof_ r0e4si6d0u3eiss (emxcoldudelinedg gulsyicninget ahned3 pQr3ol1incer,y nsotta lshsotrwunc thuerree.) Ffiitg ure4B
within the allowed regions for both the published X-ray structure and our homology model. A
shows a structural comparison between human carbonic anhydrase 1 (3LXE) and the model of
visual inspection of structures and, particularly, residues within the active site, show similarities
FusariumgraminearumcarbonicanhydraseFGSG_04603inspacefillingandskeletalformat. Askeletal
between 3LXE and modelled FGSG_04603 structures (Figure 4B). The majority of the residues in the
format compares active sites of the carbonic anhydrase proteins following structure alignment.
vicinity of the active site are identical in the two proteins, suggesting similar activity, as well as
Additioninahlilbyi,tothrye cihnahraibctietroirsti(cgsr. een),inordertomarktheactivesiteofthetwoproteins,isshown.

Figure 4. Model of Fusarium graminearum CA (FGSG_04603) in comparison to human CA. (A)
Figure 4. Model of Fusarium graminearum CA (FGSG_04603) in comparison to human CA.
Ramachadran plots showing protein backbone dihedral angles for amino acids. Dots present the
(A)Ramachadranplotsshowingproteinbackbonedihedralanglesforaminoacids.Dotspresentthe
angles (psi vs. phi) and outlined are energetically preferred angles. Shown are values for measured
angles(pansidv csa.lcpuhlai)teadn dprooutetilnin setdruactruereesn eforgr ehtuicmalalny pCrAe1f e(rmreedasaunregdle) sa.nSdh ofuwnnguasr e(cvalacluulaetsedfo);r amndea (sBu)r edand
calculatecodmppraortiesoinn sotfr ustcrtuucrteusrefso rofh uhummaann CaAnd1 f(umnegaals uCrAedI )bainnddinfgu nregguiosn(sc awlcituhl aintehdib)i;toarnyd m(Bol)eccuolmesp arison
ofstructiuncrleusdoedf hinu gmreaenn.a ndfungalCAIbindingregionswithinhibitorymoleculesincludedingreen.
Structure quality has been assessed using Ramachandran plots for the model (Figure 4A).

AcomparisonofRamachandranplotsforthe3LXEstructureandthemodelshowasimilardistribution
ofresiduesinstructuralgroupsoftheβ-sheet,aswellastheleft-andright-handedα-helix. Alarge
majority(over90%)ofresidues(excludingglycineandproline,notshownhere)fitwithintheallowed
regions for both the published X-ray structure and our homology model. A visual inspection of
structures and, particularly, residues within the active site, show similarities between 3LXE and
modelledFGSG_04603structures(Figure4B).Themajorityoftheresiduesinthevicinityoftheactive
siteareidenticalinthetwoproteins,suggestingsimilaractivity,aswellasinhibitorycharacteristics.
As only a relatively small number of RR metabolites have been experimentally tested,
other possibly stronger inhibitors may exist within the set. In order to assist the selection of
metabolites for further experimental analysis binding energy of RR metabolites have been tested
computationallyusinghigh-throughputdockinganalysisperformedusingPyRx(ScrippsResearch
Institute,LaJolla,CA,USA)[63]andAutoDock4(ScrippsResearchInstitute,LaJolla,CA,USA)[64].
Metabolites 2016, 6, 31 9 of 15
As only a relatively small number of RR metabolites have been experimentally tested, other
possibly stronger inhibitors may exist within the set. In order to assist the selection of metabolites for
Metabolites2016,6,31 9of15
further experimental analysis binding energy of RR metabolites have been tested computationally
using high-throughput docking analysis performed using PyRx (Scripps Research Institute, La Jolla,
CA, USA) [63] and AutoDock4 (Scripps Research Institute, La Jolla, CA, USA) [64]. A zinc atom has
A zinc atom has been added to the FGSG_04603 structural model according to the analysis of the
been added to the FGSG_04603 structural model according to the analysis of the structural overlap
structuraloverlapbetweentheexperimentalandmodeleddata. Chargeswerefirstcalculatedusing
between the experimental and modeled data. Charges were first calculated using the Geisteiger
theGeisteigermodel(underAutoDock4). Subsequentlythezincatomwasassignedachargeof+2and
model (under AutoDock4). Subsequently the zinc atom was assigned a charge of +2 and designated
designatedasnotchemicallybondedtotheprotein. Three-dimensionalstructuresofallligandswere
as not chemically bonded to the protein. Three-dimensional structures of all ligands were obtained
obtainedfromthePubChemdatabase[38]andwereuseddirectlyinPyRx. AutoDock4analysiswas
from the PubChem database [38] and were used directly in PyRx. AutoDock4 analysis was
performedforselectedmetabolites,includingtheassignmentofchargestoligandsusingtheGeisteiger
performed for selected metabolites, including the assignment of charges to ligands using the
modGeeli.sAteilglesrin mgoledeblo. nAdlls s(inogultes ibdoendosf r(oinugtssi)dien olfi grianngds)s iwn leirgeanrodtsa wtaebrlee rwothatialeblteh wehpirleo ttehien pwroatseikne wptasr igid.
Higkhe-tpht roriuggidh.p uHtigdho-ctkhirnogugahnpaulyt sidsowckaisngp earfnoarlmyseisd wfoarsb optehrfhourmmeadn afonrd bFoutsha rhiuummgarna maninde aFruumsarpiuromt eins.
Sevegrraamlcinoemarpumou pnrdostewinesr. eSesveelercatl ecdomfoprofuunrdths ewreteres tsinelgecbteadse fdoro fnurtthheeirr tbeisntidnign gbaesneedr gonie sthtehira tbpinrdeivnigo usly
indiecnaetergdi,eesx tpheart ipmreevnitoaullsyl,yr ienledvicaantecde,a esxapnetriifmuenngtaalllayg, erenltesv,aanscwe ealsl aansthifauvnignagl apgroenvtesn, aasc wtiveiltl yasa shainvhinibgi tors
ofpprorotevienns ainctimviatym ams ainlihainbictoerllss otfh patrohtaeivnes cilno smeaomrtmhaollioagne cseillns Fthuasta rhiauvme gclroasme ionretahroulmog.eFsu irnt hFeursatreisutmin gof
othegrrapmroinpeaorsuemd. FFuusrathrieurm tesgtrianmg ionfe oatrhuemr pCrAopionshedib Fituosrasriaurme gurnamdeinrewarauymi nCAou inrhgirboituoprs. are underway in
our group.
2.4. ExperimentalAnalysis
2.4. Experimental Analysis
Severalselectedmetabolitesfromthesetofbibliomeandcomputationally-determinedbindersto
Several selected metabolites from the set of bibliome and computationally-determined binders
CAweretestedagainstFusariumgraminearuminfungalgrowthinhibitionassays,includingtannic
to CA were tested against Fusarium graminearum in fungal growth inhibition assays, including tannic
acid,trans-Ferulicacid,naringenin,andN-(p-coumaroyl)serotonin. Theeffectofthesecompoundson
acid, trans-Ferulic acid, naringenin, and N-(p-coumaroyl) serotonin. The effect of these compounds
FusariumgraminearumgrowthinculturesisshowninFigure5withIC valueslistedinTable1.
on Fusarium graminearum growth in cultures is shown in Figure 5 with IC5050 values listed in Table 1.

Figure 5. Fungal growth inhibition assay in the presence of increasing metabolite concentrations.

(A)Radialgrowthofafungalcolonyinoculatedatthecenterofapotatodextroseagar(PDA)plate
in the presence of different metabolites after three days post inoculation; and (B) radial fungal
growth (diameter in mm) of a Fusarium graminearum inoculum measured over three days in the
presence of increasing concentrations of each metabolite. Each point shows the average of three
replicates±standarddeviation.
Metabolites2016,6,31 10of15
Table1.IC valuesfortestedcompoundsdeterminedagainstFusariumgraminearumcellcultures.
50
Compound IC (mM)
50
TannicAcid 0.58
trans-FerulicAcid 3.1
Naringenin ~1
indole-3-carboxylicacid >10
The strongest inhibitor of Fusarium graminearum growth of the tested metabolites is tannic
acid, followed by naringenin and ferlic acid. These compounds have been shown previously to
behighlypromiscuousinhibitorsofarangeofmetalloenzymesinmammaliancellsincludingCA,
P450, and lipoxygenase. The majority of protein targets for tannic and ferulic acid have close
sequenceorthologsinFusariumgraminearumandmany(e.g.,CA,P450,lipoxygenase)areimportant
in Fusarium graminearum growth and virulence. Structural analysis of these proteins has shown
significant similarity, particularly in the binding/active-site regions, suggesting activity of similar
inhibitorymolecules.
3. MaterialsandMethods
3.1. MetaboliteCharacterization
Metabolite information was obtained from automated and manual searching of all available
literaturedealingwiththesubjectofmetaboliteormetabolomicsanalysisinwheat,barley,oroats
infected with Fusarium graminearum. In addition, plant hormones, as well as metabolites listed as
generallyrelatedtobioticresponsesareincluded. Allmetabolitesdeterminedfromtheliteraturehave
beenmappedtothePubChemdatabase[38]andCIDswereassigned. Molecularcharacteristicshave
beendeterminedusingChemmineR[65],RCPI(developedbyN.Xiao),andRCDK(developedby
R. Guha), running under R and Bioconductor, and using 3D molecular structures for the analysis
ofphysiochemicalcharacteristicsofmolecules. Thesemethodsprovidedconstitutional,topological,
geometrical,andelectronicdescriptors,aswellasmolecularfingerprints,andwereusedtodetermine
overallmolecularcharacteristics,e.g.,size;drug-likenesswithLipinskirule-of-fiveproperties,and
molecularchemicalstructuresimilarities. Foracomparisonofmolecularcharacteristicswehavealso
obtainedallknownplantmetabolitesfromtheGolmdatabase[41],aswellasallFDAapproveddrugs
obtainedfromtheZINCdatabase[44].
3.2. ProteinTargetAnalysis
Experimentally-determinedproteintargetsofmetaboliteshavebeenobtainedfromthePubChem
Database [38] using Bioactivity information. Only assays with known protein targets and with
metabolites shown to be active have been retained. Several Perl and R scripts have been written
in-housetosearchforproteininformationrelatingbioassayIDswithproteintargetIDswithprotein
namesandstructures. Asmostpubliclyavailablebioassaytestinghasbeendoneforgeneralresistance
metabolites we have determined related metabolites in the selected group of resistance-related
metabolitesstudiedhereusingastructuresimilaritysearchprovidedunderChemmineR[65].
3.3. SequenceComparison
HomosapienstoFusariumgraminearumsequencecomparisonshavebeenperformedusingmethods
developed in-house. In this approach gene and protein sequences across species are compared in
ordertodeterminepossibleorthologs. Inputsequencesrepresenting13,321proteinencodinggene
models for Fusarium graminearum were obtained from the Broad Institute Fusarium Comparative
Database[56]. 89,033HumanproteinsequenceswereacquiredfromtheOctober2014releaseofthe
UniProtknowledgebase[52].