Why Argonaute is needed to make microRNA target search fast and reliable

(1)

Delft University of Technology

Why Argonaute is needed to make microRNA target search fast and reliable

Klein, Misha; Chandradoss, Stanley D.; Depken, Martin; Joo, Chirlmin

DOI

10.1016/j.semcdb.2016.05.017

Publication date

2017

Document Version

Final published version

Published in

Seminars in Cell and Developmental Biology

Citation (APA)

Klein, M., Chandradoss, S. D., Depken, M., & Joo, C. (2017). Why Argonaute is needed to make microRNA

target search fast and reliable. Seminars in Cell and Developmental Biology, 65, 20-28.

https://doi.org/10.1016/j.semcdb.2016.05.017

Important note

To cite this publication, please use the final published version (if applicable).

Please check the document version above.

Copyright

Other than for strictly personal use, it is not permitted to download, forward or distribute the text or part of it, without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license such as Creative Commons. Takedown policy

Please contact us and provide details if you believe this document breaches copyrights. We will remove access to the work immediately and investigate your claim.

This work is downloaded from Delft University of Technology.

(2)

ContentslistsavailableatScienceDirect

Seminars

in

Cell

&

Developmental

Biology

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 / s e m c d b

Review

Why

Argonaute

is

needed

to

make

microRNA

target

search

fast

and

reliable

Misha

Klein

1

_,

_Stanley

_D.

_Chandradoss

1

_,

_Martin

_Depken

∗

_,

_Chirlmin

_Joo

∗ KavliInstituteofNanoScienceandDepartmentofBioNanoScience,DelftUniversityofTechnology,Delft,2629HZ,TheNetherlands

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received15April2016

Receivedinrevisedform19May2016 Accepted21May2016

Availableonline26May2016 Keywords: Argonaute miRNA TargetSearch Lateraldiffusion Singlemolecule FRET Free-energylandscape Theoreticalmodel RNAsilencing Searchstabilityparadox MicroRNAtargetprediction

a

b

s

t

r

a

c

t

MicroRNA(miRNA)interfereswiththetranslationofcognatemessengerRNA(mRNA)byﬁnding, prefer-entiallybinding,andmarkingitfordegradation.Tofacilitatethesearchprocess,Argonaute(Ago)proteins cometogetherwithmiRNA,formingadynamicsearchcomplex.Inthisreviewweusethelanguageof free-energylandscapestodiscussrecentsingle-moleculeandhigh-resolutionstructuraldatainthelight oftheoreticalworkappropriatedfromthestudyoftranscription-factorsearch.Wesuggestthat exper-imentallyobservedinternalstatesoftheAgo-miRNAsearchcomplexmayhavetheexplicitbiological functionofspeedingupsearchwhilemaintainingspeciﬁcity.

Contents

1. Introduction...21

2. Targetsearchin1Dand3D...21

2.1. FacilitateddiffusionenablesrapidtargetsearchofmiRNA...21

2.2. Experimentalevidenceforlateraldiffusionduringtargetsearch...21

3. Multipleproteinconﬁgurationsforfastlateraldiffusionandstabletargetrecognition...21

3.1. Resolvingthespeed-stabilityparadoxbyutilizingmultiplebindingmodes...21

3.2. ExperimentalevidencefortwoinitialbindingmodesofAgo-miRNA...22

3.3. TheexperimentalevidenceforadditionalbindingmodesofAgo-miRNA...24

4. EnergylandscapeofmiRNAtargetsearch...24

5. Outlook...24

5.1. FurtherinsightintoAgo-miRNAtargetsearchcanimprovemicroRNAtargetpredictionalgorithms...24

5.2. Implicationsforothertargetsearchsystems...26

Acknowledgements ... 27

References...27

∗ Correspondingauthor.

E-mailaddresses:s.m.depken@tudelft.nl(M.Depken),c.joo@tudelft.nl(C.Joo).

1 _Co-ﬁrst_authors.

http://dx.doi.org/10.1016/j.semcdb.2016.05.017

(3)

1. Introduction

Eukaryotes regulate gene expression post-transcriptionally

throughtheRNAinterference(RNAi)pathway.Thispathwaybegins

withthetranscriptionofnon-codingRNAanditssubsequent

matu-rationintomicroRNA(miRNA).Tofacilitatesearchandsuppression

oftargetmessengerRNA(mRNA),Argonaute(Ago)proteinsjoin

togetherwiththemiRNA molecule, formingan efﬁcientsearch

complex [1,2]. In thepool of cellularRNA, thesearch complex

ﬁnds mRNA cognate to its miRNA and primes its degradation.

Asthe search relies onthermal motion, the functioning ofthe

searchcomplexcanbeunderstoodintermsofdiffusionandthe

binding-energy landscape of mRNA-Ago-miRNA interactions. In

this Review, we discuss recent single-molecule and structural

dataonAgo,and borrowfree-energyconsiderationsandtheory

fromtranscription-factorsearch,highlightinghowseveralofthe

observedAgoconformationscouldfunctiontospeedupthesearch

process.

2. Targetsearchin1Dand3D

Ever since the initial observations of an astonishingly high

associationrateoftheE.coliLacrepressortothelacoperon[3],

researchershavebeentryingtounderstandgeneralmechanisms

thatcouldspeed uptarget searchonnucleic-acidtemplates.In

theirseminalwork[4],Berg,WinterandvonHippelproposeda

facilitated diffusionmechanismby which theproteincombines

three-dimensional diffusionthroughthe cytoplasmwith lateral

diffusion along the DNA (see Fig. 1)[5]. We herequalitatively

summarizethetheoreticalargumentsbehindthissuggestionand

reviewtheexperimentalevidenceforlateraldiffusionbyvarious

searchcomplexes.

2.1. FacilitateddiffusionenablesrapidtargetsearchofmiRNA

Though facilitated diffusion was originally aimed at

transcription-factorsearchonDNA,thesameargumentsapplyto

anysearcheralonganucleicacidsequence,includingAgo-miRNA

searchonRNA.Thebeneﬁtofemployingboth3Dand1Dsearchcan

bequalitativelyunderstoodasfollows:Toﬁndthenextsequence

toprobe,itwillalwaysbefastertodiffuseashortdistancelaterally

along the RNA (through hopping and sliding; Fig. 1) than to

diffusea long distancethrough thecytosol.Aslateraldiffusion

bringsyou toclose-bysites,there existsa point beyondwhich

the search complex starts predominantly probing sites already

visited.Atthispointitbecomesfavorabletomovetoanunprobed

RNAneighborhoodbydiffusing throughthecytosol.Minimizing

redundancy of the one-dimensional(1D) search thus comesat

thecostofemployingtheslower3Dsearch,andthereexistsan

optimumpartitioningbetweenthetwo[4,6–9].

2.2. Experimentalevidenceforlateraldiffusionduringtarget

search

Single-moleculeﬂuorescencestudiesbroughtdirectevidence

oflateraldiffusionduringmoleculartargetsearch,including

slid-ingoftranscriptionfactors[10,11],DNArepairproteins[12–14]

zinc-ﬁngerproteins[15],andtheDNArecombinationproteinRecA

[16]. Like Argonaute, RecA makes a nucleoprotein complex (a

RecA—single-strandedDNAﬁlament)thatisreadytobasepairfor

targetsearch[17–21].Inordertoinvestigatelateraldiffusionof

Ago-miRNAonRNA,weadoptedaninvitrosingle-moleculeFRET

assaythatwasdevelopedforstudyingRecA-mediatedtargetsearch

[16].WeplacedtwoidenticalbindingsitesonasingletargetRNA

strand,eachofwhichledtoadifferentFRETefﬁciencywith

Ago-miRNAbound[22].Weobservedthatasubstantialfractionofthe

Fig.1. Facilitateddiffusion.

Fourdifferentmodesofsearchcaninprinciplebedistinguished.1)3Dsearch:An Argonauteproteinprobesanewsequencebyfirstunbinding,thendiffusingthrough thecytosol,andfinallybindingtoprobeanewuncorrelatedsite.2)Sliding:A non-specificallyboundproteinlaterallydiffusesalongthemRNAtoprobeanewsite, probingeverypotentialintermediatesitefromthestarttothenewsite.3)Hopping: Anon-specificallyboundproteinunbinds,butquicklyrebindsagaintoasiteclose by(alongtheRNA)fromwhereitunbound,butnotnecessarilyprobingeverysitein between.4)Intersegmentaltransfer:ahoppingmechanismwhereunbindingand bindingpositionsarecorrelatedin3Dspace,butfarapartalongtheRNA.Thisis possibleduetothecoiledconformationRNAadaptsinvivo.

bindingevents(>50%)shuttledbetweentwostrongbinding

posi-tionsviarapidlateraldiffusion.Whenusingavolume-occupying

reagent (PEG) tomimic physiological conditions, most binding

events(>90%)displayedshuttlingbythesameAgo-miRNA

com-plex.Thissuggeststhatlateraldiffusioncouldalsobeimportant

forinvivomicroRNAsearch.

3. Multipleproteinconﬁgurationsforfastlateraldiffusion andstabletargetrecognition

Whiletargetsearchisspedupbyfacilitateddiffusion,Slutsky

andMirny[8,23]arguedthatitisnotpossibletohavebothfast

lateraldiffusionandstable/preferentialbindingtothetargetusing

asinglenucleoproteinconformation.Themorestablebindingto

thetargetis, themorestablebinding tosimilarsequences also

becomes,andthelateraldiffusionslowsdownasitgets

increas-inglytrappedatnon-targetsites.Tounderstandwhatisneeded

fortheresolutionofthisapparentparadox,wenowfollowSlutsky

andMirny[8,23]andconsiderthestatisticalvariationofbinding

energiesalongthesubstrate(whichforusismRNA).

3.1. Resolvingthespeed-stabilityparadoxbyutilizingmultiple

bindingmodes

Apartfromthetarget,thesequencesbeingsearchedthrough

canbeconsideredasessentiallyrandomanduncorrelated[8,24].

Asubstantiallypreferentialbindingtothetargetrequiresthata

correctmatchhasaconsiderableenergeticdifference(E,for

deﬁ-nitionseeFig.2A)toallpartialmatches.SlutskyandMirnyassume

thatthesearchcomplexhasabindingenergyroughlyproportional

tothedegreeofsequencehomologybetweenprobedandtarget

sequence.Undertheassumptionthatthebindingenergycomes

onlyfromindividualnucleotide-basepairingenergies,alarge

ener-geticdifferencebetweentargetandnon-targetpositionscanonly

beachievedbylargedifferencesinpairingforeachnucleotide.A

generalincreaseofbasepairingenergiesresultsinalargerstandard

deviationamongbindingenergiesatdifferentpositions(compare

Rofthe“recognition”landscapeandSofthe“search”landscape

inFigs.2BandC respectively),andthediffusionconstantalong

themRNAcanbeshowntodecreasesharply[8,25].InFig.2Bwe

(4)

Fig.2.Search-stabilityparadox.

(A)Energiesofthebindingsitesareshownasshortblackhorizontalmarkers.Being asumofbasepairingenergies,bindingenergiesare(approximately)Gaussian dis-tributedwithastandarddeviation.Thetargetsiteisseparatedfromtheother bindingsitesbyanenergyofaboutE.Whendiffusinglaterally,theminimal bar-riertowardsdiffusionissetbytheenergeticdifferencebetweenneighbouringsites (E†_)._In_reality_there_are_intervening_barriers,_as_depicted_by_the_dashed_line._With littlelossofgenerality,wewillignoretheseadditionalcontributionstothebarriers andfocusonthebest-casescenario.

(B)Recognitionmode−Stablebinding,butslowsearch:Alargerdifferencebetween targetandnon-targetenergiescomesatthecostofhavinglargerbarrierstowards diffusion.Therightpanelshowsthecompletedistributionofenergeticstates (stan-darddeviationR)ofwhichasubsetisplottedintheleftpanel.Thetypical(minimal)

barriertowardsdiffusion(E†

R)anddifferentialbindingenergy(ER)areindicated.

(C)Searchmode−Fastsearch,butnostablebinding:Decreasingthebarriersalso decreasesthedifferencebetweentargetandnon-targetenergy,whichhampersthe abilityofthesearchcomplextoselectivelybindtothetarget.Therightpanelshows thecompletedistributionofenergeticstates(standarddeviationS)ofwhicha

sub-setisplottedintheleftpanel.Thetypical(minimal)barriertowardsdiffusion(E†

S)

anddifferentialbindingenergy(ES)areindicated.

barrierstolateraldiffusion(fordeﬁnitionseeFig.2A),resulting

inaslowsearchprocess.Reversely,inFig.2Cweillustratehow

smallbarrierstodiffusionimpliespoorrecognition.Slutskyand

Mirnyproposedthatthecouplingbetweenrecognitionenergyand

diffusionbarrier(E†_being_proportional_to_E)_can_be_broken_if

thesearchcomplexcanstochasticallyswitchbetweentwointernal

modeswithdifferentbindingenergystrength(Fig.2D):

Asearch(S)mode:smallafﬁnitydifferencesandfastdiffusion

(_␴S<

∼2kBTRef.[8])

Arecognition(R)mode:largeafﬁnitydifferencesandslow

dif-fusion(␴R>

∼5kBTRef.[8])

Anefﬁcientsearchermusthaveevolvedtheabilitytocombine

thesearchandrecognitionmodes.Thereby,thenon-speciﬁc

(aver-age)energies(dashedlinesinFig.2B–D)arearrangedsuchthatall

energiesofthesearchmodeliebetweentheenergiesofall

non-targetsitesandthetargetintherecognitionmode(seeFig.2D).

Suchsystemspredominantlymoveaccordingtothesearchmode

whennotatthetargetsite,butpredominantlyoccupythe

recog-nition modeonceat thetarget (see states withorange dotsin

Fig.2D).Theeffectivesearchbarriersarenowsetbythesearch

mode(E†_≈_E†

S)whiletherecognitionenergiesaresetbythe

recognitionmode(E≈ER).Bothfastsearchandstable

recogni-tionisthusinprinciplepossibleifthesearchingproteinpossessesat

leasttwodistinctbindingmodes,andtheabovecaserepresentsthe

theoreticalidealscenario(formoregeneralcasessee[6,26–29]).

3.2. Experimentalevidencefortwoinitialbindingmodesof

Ago-miRNA

Bothrecentstructuralandsingle-moleculedataofeukaryotic

Agoproteinssuggestthatthehybridizationbetweenguideand

tar-getisgradualandiscoupledtostructuralchangesinthesearch

complex.Weherediscussthesestudiesinthelightofa

search-stabilityparadoxforAgo-miRNA.

Biochemical, structural and computational analyses suggest

that Argonaute divides its miRNA guides into ﬁve functional

domains(5anchor,seed,midregion,3supplementaryregion,and

thetail region) (Fig. 3).The seed region(nt 2–8) is crucial for

genesuppression[1,17–19,30–32],anditwasshownthatprotein

mediatedinteractionsstabilizent2–6intoanA-form-helixthat

exposesnt2–4(or2–5)forbaseparingwiththetarget(Fig.4A)[33].

Basedonthisobservation,Schirleetal.[33]proposedastep-wise

targetrecognitionforhumanArgonaute-2(hAgo2),inwhichthe

initialrecognitionofthetargetoccursinthe5partofthemiRNA.

Twosubsequentsingle-moleculestudiesshowedthatAgo-miRNA

indeedusesthis so-calledsub-seedfortheinitialweak

recogni-tion.Solomonetal.designeddi-nucleotidemutationconstructsfor

mouseAgo-miRNAandmeasuredtheunbindingratefromthe

tar-getRNA[34].Wehavealsoshownthat,whenthepairedregionwas

graduallyshrunkfromthefullseed(nt2–8)toonlytheﬁrstthree

nucleotides(nt2–4),nodifferenceinthebindingratewas

notice-able[22].Thesetworesultsshowedthatitisonlytheﬁrstthree

nucleotidesoftheseedthatareusedtomaintainweakinteraction

duringtheinitialsearch.

Thetwosingle-moleculeworksalsosuggestedthatAgo-miRNA

exhibitsasharpincreaseinthebindingafﬁnitywhenthenumber

ofpairednucleotideschangesfrom6to7[22,34].Comparisonof

crystalstructuressuggeststhatthispropertyoriginatesfromthe

factthatArgonautemakestheguidekinkawayfromtheA-form

(D)Search+Recognition:Fastsearchandstablebinding:Ifthesearchcomplex pos-sesses(atleast)twodistinctbindingmodes,itbecomespossibletocombinethe landscapesofﬁguresB(blue)andC(green)toenablerapiddiffusion

E†_≈_E†

S

(5)

Fig.3. StructuralanddomainoverviewofhAgo2andmiRNA.

(A)ThebinarystructureofhAgo2-miRNAshowingfourwellconserveddomainsamongArgonauteproteins(snapshotofthestructure4W5Ntakeninpymol).

(B)ArgonauteproteinsdividemiRNA(orange)intoseveraldomains.The5_phosphate_and_nt₁_of_miRNA_(anchor)_is_bound_to_the_pocket_in_the_MID_domain._The_nt_2–8_are

knownasseedsequence,astheyarecrucialforinitialtargeting.Thent9–10havetheleastsigniﬁcantroleintargetrecognitionandareknownasthemidregion.The3

supplementaryregioniscomprisedofnt13–16,theyalsohaveconsiderableroleinstabilizingmiRNA-targetinteraction.Thenucleotidesbeyondthe16thdonotbasepair withthetargetandarecalledthetailregion.The3OHisboundtothebindingpocketinPAZdomainmakingitasa3anchor.Thet1Adenosine(t1A)inthetargetRNA(pink) bindstothebindingpocketinMIDdomain.

Fig.4. SeedofmiRNAandhAgo2-helix7.

(A)Nucleotides2–4(green)oftheguideRNAarewellexposedbyresiduesinthePIWIdomain(goldensurface)possiblyforinitialtargetrecognition(snapshotofthestructure 4W5Ntakeninpymol).

(B)Theaccesstont5–7oftheguide(green)isblockedbythehelix-7motif(red).Thebaseparingoftargettoguident5–7wouldrequiredisplacementofhelix-7(snapshot ofthestructure4W5Ntakeninpymol).

(C)Uponbaseparingwiththetarget(grey)thehelix-7motifisdisplacedby4Åcomparedtoguide-onlystructure.Thedisplacementofhelix-7removestheconstraintsfrom nt6and7(yellow)comparedtoguideonlystructure(green)makingnt6and7availableforbaseparing(seetheclose-upviewintherightpanel).(snapshotofthestructures 4W5N(guideonly)and4W5O(guideandtarget)takeninpymol).

(6)

stackedstructureinseveralplaces[17–20].Themostprominent

kinkdisruptingthehelicalarrangementoftheguideisbetweennt

6and7(Fig.4B).Baseparingtothetarget,therefore,requiresashift

ofthehelix-7thatclasheswiththeincomingtarget.Afterpairing

ofnt2–4,hAgo2undergoesaconformationalchangeleadingtoa

4Ådisplacementofthehelix-7loopandallowingbasepairingof

nt6–8(Fig.4C).Itwashypothesizedthatthesharpincreaseinthe

timeboundbetweenhaving6and7ntmatchingiscausedbythe

conformationalchangeofthehelix-7motif[33].Weheresuggest

thatAgomakesachangefromaweakbinding(search)modeusing

nt2–4toa strongbinding(recognition)modeusinga fullseed

throughtheconformationalchangeofthehelix-7.

3.3. Theexperimentalevidenceforadditionalbindingmodesof

Ago-miRNA

In addition to the helix-7 movement, more conformational

changestakeplaceafterseedpairingisachieved,andbeforethe

boundAgo-miRNAcomplexbecomescleavagecompetent.

First,bindingofthesupplementaryregion(nt13–16)ensuing

theseedpairingenhancesthebindingstabilityofAgo-miRNA[34].

Butthepairingbeyondnt8isrestrictedbyaphysicalconstraint

[33](Fig.5A).

WideningupofachannelbetweenPAZandN-terminusdomains

allowsforarearrangementofthedisorderedsupplementaryregion

(nt13–16)ofthemiRNAintoahelicalA-form,preparingitfor

pair-ingwiththetargetRNA(Fig.5B)[33].Itremainstobeseenwhether

targetrecognitionisenhancedbythisadditionalcheckpoint.

Second,biochemicalandsingle-moleculestudieshaveshown

thatthebaseparinginthemidregionisnecessaryforcleavage

oftargetRNA[32,35].ButJoetal.alsoobservedthatasigniﬁcant

portionofAgo-miRNAswasnotabletocleavethetargetRNAsin

spiteoftheirperfectcomplementarity[32,35].Theunsuccessful

cleavageofperfectcomplementarytargetmightbetheresultant

ofafailuretoinduceanadditionalconformationalchangeneeded

forcleavagethatinvolvespositioningofAgo’scatalyticresidues

residingnearnt9–10ofthemiRNA.

Third,AgousesitsPAZdomaintoprecludemiRNAfrombeing

tightlyassociatedwithtargetRNA.Anearlierbiochemicalstudy

reportedthatbareRNAasshortas12bpislongenoughforstable

hybridization(∼ayearoflifetime)[36].Butitwasobservedthat

Ago-miRNA(orAgo-guideDNA)oftendissociatedfromitstarget

withinsecondstominutesafterbinding[32].Thisreversible

bind-ing,which isspeculatedtoreduceoff-targeting[37],ispossible

becausethe3endofguideRNAisanchoredtothePAZdomainand

thislowersthebindingafﬁnityofAgo-miRNA(especiallyatthe3

end)totargetRNA[20,37–41].

InadditiontothecomplexinteractionsbetweenAgoandaguide

strand,adirectinteractionbetweenAgoandtargetRNAalso

con-tributestothetargetselection.Schirleetal.[42]showedthathAgo2

interactswiththeadeninenucleotideofthetargetwhenitis

oppo-sitetothe1stnucleotideoftheguide.Throughawaternetwork,the

residuesintheMIDdomain(Fig.3A)speciﬁcallyrecognizethet1A

anchoringtheAgo-miRNAcomplextothetarget.Usinga

single-moleculeassaytheyshowedthatt1A doesnotinﬂuenceinitial

targetrecognitionbutincreasestheresidencetimeofAgo-miRNA

ontothetargetRNA,whichmightenhanceitscleavageefﬁciency

[42].

4. EnergylandscapeofmiRNAtargetsearch

Havingdiscussedtheevidencethataseriesofconformational

changesareneededtoinitiatestablebindingandcleavageof

tar-getmRNA,wenowdiscusshowconformationalchangeseffectthe

binding-energylandscape.WhenAgoinitiallyscansthetargetRNA

itexposesonlynucleotides2–4ofthemiRNA,termedthesub-seed.

InthissearchmodeitdoesnotdiscriminatestronglybasedonRNA

sequence,andlateraldiffusionislikelyrapid.Acompletematchof

thesub-seedstabilizesaconformationalchangethatexposesthe

remainderoftheseed(nt2–8)forbasepairing,and,oncepaired,it

slowsdownthediffusioninthisrecognitionmode(Fig.6A).Upon

encounteringasequence bearingcomplementaritytotheentire

seed,thehelix-7is displacedtoallowmiRNAtofullypairwith

thetarget,andtheAgo-miRNAcomplexarrivesinthismorestable

recognitionstate(Fig.6AandB).Wesuggestthatthefunctionof

thesevariousstatesisanalogoustothefunctionofinternalstates

intranscription-factorsearch(Fig.2D).

InFig.6Bwesketchafree-energylandscapeofthedominant

conﬁgurationatvaryingdegreesofbasepairingforaperfectmatch.

Transitionsrequiringconformationalchangescostenergy,

increas-ingbarrierstofurtherbasepairing.Weconstructasketchofthe

landscapebasedonasingle-moleculestudythatreportedthe

exis-tenceofvariouspathwaysevenwhenthefullsequenceofmiRNA

matcheswitha target[35]:a signiﬁcantfractionofthe

popula-tionshowedtransientbinding(∼10%)andstablebindingwithno

cleavage(∼30%).

Assumingthatthelargestbarriertofurtherbasepairing

orig-inates from the required movement of helix-7,the substantial

fractionoftransientlybindingproteinsindicatesthatthisbarrier

mustcomeclosetothebarriertounbind.Further,theevenlarger

fractionofstablebutnon-cleavingcomplexesindicatesthatthe

averagebindingenergypasthelix-7isstrong,andthatthe

cleav-agerateisslowcomparedtoexperimentaltimes,butfastcompared

tounbinding.

Withthesegeneralconsiderations,weconcludethatthe

free-energy landscape of Fig.6B capturesat least one search mode

(pre-seedpairing) andatleastonerecognitionmode(post-seed

pairing).Thesetwomodescouldbefurthersplitup,e.g.theseed

pairingintosub-seedandfullseedpairing.Still,thegeneral

prin-ciplebehindresolvingthespeed-stabilityparadoxshouldapply.

To determine thequantitative effects of this energy landscape

will requireadditional theoretical work accounting for gradual

basepairingandaseriesofconformationalchanges.Using

single-moleculetechniquesandhighresolutionstructuralstudies,itwill

alsobepossibletotesttheeffectofAgo’sconformationalchanges

ontargetsearchbyanalysingmutatedproteinsordirectlyobserve

conformationalswitching(forinstancebyusingFRETsuchasdone

forCas9in[43]).

5. Outlook

Wehavereviewedtheprinciplesbehindfacilitateddiffusionand

thespeed-stabilityparadoxingeneraltargetsearchprocesses,as

wellastheexperimentalevidenceforfacilitateddiffusioninmiRNA

targetsearch.Wefurtherdiscussedtheevidenceformultiplesearch

statesintheAgo-miRNAsearchcomplex,whichcouldhelpresolve

thespeed-stabilityparadox—simultaneouslyenabling thesearch

tobefastandthebindingtothetargettobestrong.

5.1. FurtherinsightintoAgo-miRNAtargetsearchcanimprove

microRNAtargetpredictionalgorithms

DuetothecomplexnatureofthemRNAtargetingprocess,itis

farfromstraightforwardtopredictwhatgenesaresilencedbya

particularmiRNA.Experimentally,mRNAtargetshavebeenfound

byanalysingtheeffectofmiRNAexpressiononprotein

produc-tionorbyperformingbindingassays[44].Forsuchapproachesto

work,oneneedstoknowwhattargetgeneshouldbeconsidered

(7)

Fig.5. Cleavagecompetentstate.

(A)Structureshowingthebasepairingbetweenaguidestrand(green)andatargetstrand(red).Thebasepairingbeyondnt8(g8)isblockedbyaresidueF811inahelixof thePIWIdomain(snapshotofthestructure4W5Otakeninpymol).

(B)AbinarystructureofhAgo2-miRNAshowingthedisordered3_{supplementary}_region_of_guide_RNA_(green)_passing_through_a_channel_between_N_domain_(blue)_and_PAZ

domain(purple)(snapshotofthestructure4W5Ntakeninpymol).

(C)AternarystructureofhAgo2-miRNAanditstargetshowinganA-formhelicalarrangementofthe3_{supplementary}_region_of_guide_(green)_in_ternary_structure_(snapshot

ofthestructure4W5Otakeninpymol).

sitesarescored,andhighscoringtargetsaresubsequentlytested

inexperiment.

SimplesequencehomologybetweenthemRNAtotheguiding

miRNAdoesnotbyitselfgiveanaccuratepredictionof targets.

Presently,typicalpredictionalgorithmsarelargely

phenomeno-logicalinnature,forexample,assigninghigherscorestosequences

thatfullymatchtheseedofthemiRNAand/orareevolutionary

conserved.Additionally,accountingforthesecondarystructureof

mRNAandthesequence outside ofthetargeted3-UTRfurther

improvespredictions[44,45].Arecent combinedbioinformatics

and invivo study showedthat there areat least 14 additional

sequencefeatures(forexamplethelength3-UTRregionandthe

predictedstructuralaccessibilityof theRNA)ofthemRNAthat

improvemicroRNAtargetpredictionalgorithms[46].Yet,despite

mucheffort,predictionalgorithmsoftenpointtomanytargetsites

thatcannotbevalidatedexperimentallyorfailtopickouttargets

thathavebeenpreviouslyvalidated.

Single-moleculestudiesallowonetostudyhowAgo-miRNA’s

interactionwithRNAbindingproteinseffectstargetafﬁnity.

Syn-thesisingsuchmolecularlevelunderstandingintothefree-energy

landscapes that we have discussed in this review should help

improvingthescoringfunctionsoftargetpredictionalgorithmsby

takingthenon-equilibriumfeaturesofthesystemintoaccount.

Additionally,predictionalgorithmscanpotentiallybeimprovedby

takingsequencesneighbouringthetargetintoaccount[30,47–49].

(8)

Sub-seed pairing nt 2-4 Seed pairing nt 2-8 (stabilized by helix-7) Cleavage competent state Pairing nt 2-7 5’ 3’ PAZ MID PIWI N helix-7 miRNA sub-seed pairing (nt 2-4) nt 2-7 pairing (Search mode)

Post seed pairing (Recognition mode)

Cleavage

Extent of base pairing (starting from nt 2) Free-energy helix-7 (from nt 2-7 to nt 2-8) Solution state

A

B

Fig.6.TargetsearchprocessbyhAgo2.

(A)AmodelsummarizingconformationalchangesduringtargetsearchbyhAgo2-miRNA.Inlightofthesearch-stabilityparadoxdiscussedinFig.2weidentifyatwosearch modes(pink+green)andarecognitionmode(blue).Alternatingbetweensearchandrecognitionmodesisenabledthroughthemovementofthehelix-7motif(orange). (B)Schematicfree-energydiagramforAgo-microRNAtargetrecognition.Formingbondsbetweentargetandguide(horizontalaxis)makesthecomplexmorestable(vertical axis).Inlightofthesearch-stabilityparadox,asproposedbySlutskyandMirnyanddiscussedinFig.2,weidentifyatleast1searchmode(pre-seedpairing,greenarrow) andatleastonerecognitionmode(post-seedpairing,bluearrow).Toresolvetheparadox,Argonautecanusethemovementofitshelix-7motiftoswitchbetweensearch andrecognitionmodes(orangearrow).Potentially,additionalmodescanbedistinguished,suchassub-seedpairing(pinkarrow).

neighbouringeachother,thetotalretentiontimewassubstantially

largerthanwhatcanbeexpectedontheoreticalgroundsfortwo

non-interactingtargets[22].Thissynergisticeffectmightalsobe

observedwhenatargetisneighbouredbysub-seedsequences.It

willbeinterestingtodeterminewhetherthisputativeeffectexists

invivo.Possibly,modellingthephysicalinteractionwith

neigh-bouringsites, and accordingly assigning higher scoresto those

mRNAsequenceswithahigh-densityofsub-seedsequences,could

thenimprovetargetpredictionalgorithms.

5.2. Implicationsforothertargetsearchsystems

Inthecell,multiplenucleicacid-mediatedtargetsearch

pro-cessestakeplace.Amongthem,RecA-mediatedtargetsearchisthe

mostthoroughlystudiedsystem.Qietal.[50]selectivelyobserved

stableinteractionsbetweenaRecA-ssDNAhomologueandDNAin

aDNAcurtainexperiment,inwhichsingle-moleculesignalswere

onlyobservedwhenssDNAanddsDNAmatchedwitheachother

foratleast8nucleotides.Furthermore,usingsingle-moleculeFRET,

Ragunathanetal.[16]observedshort-livedinteractions(1–10s)

between RecA-ssDNA and target DNA that had 5–7 matching

nucleotides.Thedifferencebetweenhaving7or8matchessuggests

thereexistsaratelimitingstephamperingRecA-ssDNAﬁlaments

toextendbasepairingbeyondthe7thnucleotide(similartothe

barrierrepresentingthemovementofthehelix-7motifinFig.6B).

Recently,greatattentionhasbeenbroughttotheCRISPR/Cas

system,anadaptiveimmunesysteminbacteria,whichusesRNA

asaguidetotargetforeignDNAorRNA[51].CRISPR’stargetsearch

involvesaprotein-DNAinteraction(recognitionofa3-ntsequence,

so-calledPAMsequence)andRNA-DNAinteractions.Biochemical

studiessuggestedthatitisthePAMrecognitionthatoccurspriorto

theseedrecognition[52–54].Recently,astructuralstudyshowed

thattheﬁrst8nucleotidesofCas9sguidearepre-organizedina

helicalA-form,similartotheseedsequenceofmicroRNAin

Arg-onaute[55].ArecentFRETstudyindicatedthatthereisanother

modethatfollowsbindingoftheseedrecognition[43].Theauthors

showedthatonlywhentheguideRNAofCas9makesextensivebase

pairing(∼16ntoutofthe20ntguide),anucleasedomain(HNH)

migratestowardsthetargetDNA.Altogether,theﬁndingsimply

thatCRISPR/Cas9,similartoArgonaute,usesmorethantwo

bind-ingmodestoovercomethespeed-stabilityparadox(‘PAMonly’

to‘PAM+seed’to‘cleavagecompetent’).WhereasaDNAcurtain

assayruledoutlongdistancelateraldiffusion,itwillbeinteresting

(9)

lat-eralexcursionswhensearchingforthePAMsequence.Similarly,

nolargescalelateraldiffusionhasbeenobservedforRecA/Rad51

systemsusingDNAcurtainassays(>100nmresolution)[56],while

short-rangelateraldiffusionwasobservedinsingle-moleculeFRET

experiments(nanometerresolution)[16].

Finally,itwillbeinterestingtoﬁndouthowmuchthesearch

mechanism of human Argonaute-2 is shared withother target

searchsystemssuchasthosementionedinthis reviewand

dif-ferentclassesofAgoproteinsthatuseDNAtotargetDNA[57,58]

andRNAtotargetDNA[59]aswellasPIWIproteins[60].

Acknowledgements

C.J. was funded by European Research Council under the

European Union’s Seventh Framework Programme

[FP7/2007-2013]/ERCgrantagreementno[309509].M.D.wassupportedbya

TUDelftstartupgrant.ThisworkwassupportedbytheNetherlands

OrganizationforScientiﬁc Research(NWO/OCW),aspartofthe

FrontiersinNanoscienceprogram.WethankAafkevandenBerg

andTaoJu(Thijs)Cuiforcriticallyreadingandcommentingonthe

manuscript.

References

[1]D.P.Bartel,MicroRNAs:targetrecognitionandregulatoryfunctions,Cell136 (2009)215–233.

[2]V.N.Kim,J.Han,M.C.Siomi,BiogenesisofsmallRNAsinanimals,Nat.Rev. Mol.CellBiol.10(2009)126–139.

[3]A.D.Riggs,S.Bourgeois,M.Cohn,Thelacrepressor-operatorinteraction3. Kineticstudies,J.Mol.Biol.53(1970)401–417.

[4]O.G.Berg,R.B.Winter,P.H.vonHippel,Diffusion-drivenmechanismsof proteintranslocationonnucleicacids1.Modelsandtheory,Biochemistry20 (1981)6929–6948.

[5]P.H.vonHippel,O.G.Berg,Facilitatedtargetlocationinbiological-systems,J. Biol.Chem.264(1989)675–678.

[6]M.Bauer,R.Metzler,GeneralizedfacilitateddiffusionmodelforDNA-binding proteinswithsearchandrecognitionstates,Biophys.J.102(2012)2321–2330.

[7]M.Coppey,O.Bénichou,R.Voituriez,M.Moreau,Kineticsoftargetsite localizationofaproteinonDNA:astochasticapproach,Biophys.J.87(2004) 1640–1649.

[8]M.Slutsky,L.A.Mirny,Kineticsofprotein-DNAinteraction:facilitatedtarget locationinsequence-dependentpotential,Biophys.J.87(2004)4021–4035.

[9]H.-X.Zhou,RapidsearchforspeciﬁcsitesonDNAthroughconformational switchofnonspeciﬁcallyboundproteins,Proc.Natl.Acad.Sci.U.S.A.108 (2011)8651–8656.

[10]P.Hammar,P.Leroy,A.Mahmutovic,E.G.Marklund,O.G.Berg,J.Elf,Thelac repressordisplaysfacilitateddiffusioninlivingcells,Science336(2012) 1595–1598.

[11]A.Tafvizi,F.Huang,A.R.Fersht,L.A.Mirny,A.M.vanOijen,Asingle-molecule characterizationofp53searchonDNA,Proc.Natl.Acad.Sci.U.S.A.108 (2011)563–568.

[12]P.C.Blainey,A.M.vanOijen,A.Banerjee,G.L.Verdine,X.S.Xie,Abase-excision DNA-repairproteinﬁndsintrahelicallesionbasesbyfastslidingincontact withDNA,Proc.Natl.Acad.Sci.U.S.A.103(2006)5752–5757.

[13]J.Gorman,A.Chowdhury,J.A.Surtees,J.Shimada,D.R.Reichman,E.Alani,E.C. Greene,Dynamicbasisforone-dimensionalDNAscanningbythemismatch repaircomplexMsh2-Msh6,Mol.Cell28(2007)359–370.

[14]C.Jeong,W.K.Cho,K.M.Song,C.Cook,T.Y.Yoon,C.Ban,R.Fishel,J.B.Lee, MutSswitchesbetweentwofundamentallydistinctclampsduringmismatch repair,Nat.Struct.Mol.Biol.18(2011)379–385.

[15]L.Zandarashvili,A.Esadze,D.Vuzman,C.A.Kemme,Y.Levy,J.Iwahara, BalancingbetweenafﬁnityandspeedintargetDNAsearchbyzinc-ﬁnger proteinsviamodulationofdynamicconformationalensemble,Proc.Natl. Acad.Sci.U.S.A.112(2015)E5142–5149.

[16]K.Ragunathan,C.Liu,T.Ha,RecAﬁlamentslidingonDNAfacilitates homologysearch,Elife1(2012)e00067.

[17]E.Elkayam,C.D.Kuhn,A.Tocilj,A.D.Haase,E.M.Greene,G.J.Hannon,L. Joshua-Tor,Thestructureofhumanargonaute-2incomplexwithmiR-20a, Cell150(2012)100–110.

[18]K.Nakanishi,D.E.Weinberg,D.P.Bartel,D.J.Patel,Structureofyeast argonautewithguideRNA,Nature486(2012)368.

[19]N.T.Schirle,I.J.MacRae,Thecrystalstructureofhumanargonaute2,Science 336(2012)1037–1040.

[20]Y.Wang,G.Sheng,S.Juranek,T.Tuschl,D.J.Patel,Structureofthe guide-strand-containingargonautesilencingcomplex,Nature456(2008) 209–213.

[21]Z.Chen,H.Yang,N.P.Pavletich,Mechanismofhomologousrecombination fromtheRecA-ssDNA/dsDNAstructures,Nature453(2008)484–489.

[22]S.D.Chandradoss,N.T.Schirle,M.Szczepaniak,I.J.MacRae,C.Joo,Adynamic searchprocessunderliesMicroRNAtargeting,Cell162(2015)96–107.

[23]L.A.Mirny,M.Slutsky,Z.Wunderlich,A.Tafvizi,J.S.Leith,A.Kosmrlj,Howa proteinsearchesforitssiteonDNA:themechanismoffacilitateddiffusion,J. Phys.AMathTheor.42(2009)434013.

[24]U.Gerland,J.D.Moroz,T.Hwa,Physicalconstraintsandfunctional

characteristicsoftranscriptionfactor-DNAinteraction,Proc.Natl.Acad.Sci.U. S.A.99(2002)12015–12020.

[25]R.Zwanzig,Diffusioninaroughpotential,Proc.Natl.Acad.Sci.85(1988) 2029–2030.

[26]O.Bénichou,Y.Kafri,M.Sheinman,R.Voituriez,Searchingfastforatargeton DNAwithoutfallingtotraps,Phys.Rev.Lett.103(2009)1–4.

[27]R.Murugan,Theoryofsite-speciﬁcDNA-proteininteractionsinthepresence ofconformationalﬂuctuationsofDNAbindingdomains,Biophys.J.99(2010) 353–359.

[28]J.Reingruber,D.Holcman,TranscriptionfactorsearchforaDNApromoterina three-statemodel,Phys.Rev.EStat.Nonlin.SoftMatterPhys.84(2011)1–4.

[29]S.Yu,S.Wang,R.G.Larson,ProteinssearchingfortheirtargetonDNAby one-dimensionaldiffusion:overcomingthespeed-stabilityparadox,J.Biol. Phys.39(2013)565–586.

[30]S.L.Ameres,J.Martinez,R.Schroeder,MolecularbasisfortargetRNA recognitionandcleavagebyhumanRISC,Cell130(2007)101–112.

[31]M.Khorshid,J.Hausser,M.Zavolan,E.vanNimwegen,Abiophysical miRNA-mRNAinteractionmodelinferscanonicalandnoncanonicaltargets, Nat.Methods10(2013)253–255.

[32]L.M.Wee,C.F.Flores-Jasso,W.E.Salomon,P.D.Zamore,Argonautedividesits RNAguideintodomainswithdistinctfunctionsandRNA-bindingproperties, Cell151(2012)1055–1067.

[33]N.T.Schirle,J.Sheu-Gruttadauria,I.J.MacRae,StructuralbasisformicroRNA targeting,Science346(2014)608–613.

[34]W.E.Salomon,S.M.Jolly,M.J.Moore,P.D.Zamore,V.Serebrov, Single-moleculeimagingrevealsthatargonautereshapesthebinding propertiesofitsnucleicacidguides,Cell162(2015)84–95.

[35]M.H.Jo,S.Shin,S.R.Jung,E.Kim,J.J.Song,S.Hohng,Humanargonaute2has diversereactionpathwaysontargetRNAs,Mol.Cell59(2015)117–124.

[36]D.Herschlag,Implicationsofribozymekineticsfortargetingthecleavageof speciﬁcRNAmoleculesinvivo:moreisn’talwaysbetter,Proc.Natl.Acad.Sci. U.S.A.88(1991)6921–6925.

[37]S.R.Jung,E.Kim,W.Hwang,S.Shin,J.J.Song,S.Hohng,Dynamicanchoringof the3’-endoftheguidestrandcontrolsthetargetdissociationof

argonaute-guidecomplex,J.Am.Chem.Soc.135(2013)16865–16871.

[38]A.Deerberg,S.Willkomm,T.Restle,Minimalmechanisticmodelof siRNA-dependenttargetRNAslicingbyrecombinanthumanargonaute2 protein,Proc.Natl.Acad.Sci.U.S.A.110(2013)17850–17855.

[39]H.M.Sasaki,Y.Tomari,ThetruecoreofRNAsilencingrevealed,Nat.Struct. Mol.Biol.19(2012)657–660.

[40]A.Zander,P.Holzmeister,D.Klose,P.Tinnefeld,D.Grohmann,

Single-moleculeFRETsupportsthetwo-statemodelofargonauteaction,RNA Biol.11(2014)45–56.

[41]T.Kawamata,Y.Tomari,MakingRISC,TrendsBiochem.Sci.35(2010) 368–376.

[42]N.T.Schirle,J.Sheu-Gruttadauria,S.D.Chandradoss,C.Joo,I.J.MacRae, Water-mediatedrecognitionoft1-adenosineanchorsargonaute2to microRNAtargets,Elife4(2015).

[43]S.H.Sternberg,B.LaFrance,M.Kaplan,J.A.Doudna,Conformationalcontrolof DNAtargetcleavagebyCRISPR-Cas9,Nature527(2015)1–14.

[44]M.Thomas,J.Lieberman,A.Lal,DesperatelyseekingmicroRNAtargets,Nat. Struct.Mol.Biol.17(2010)1169–1174.

[45]D.P.Bartel,http://www.targetscan.org/.TargetScan7.0.(2015). [46]V.Agarwal,G.W.Bell,J.W.Nam,D.P.Bartel,PredictingeffectivemicroRNA

targetsitesinmammalianmRNAs,Elife4(2015).

[47]J.A.Broderick,W.E.Salomon,S.P.Ryder,N.Aronin,P.D.Zamore,Argonaute proteinidentityandpairinggeometrydeterminecooperativityinmammalian RNAsilencing,RNA17(2011)1858–1869.

[48]A.Grimson,K.K.Farh,W.K.Johnston,P.Garrett-Engele,L.P.Lim,D.P.Bartel, MicroRNAtargetingspeciﬁcityinmammals:determinantsbeyondseed pairing,Mol.Cell27(2007)91–105.

[49]P.Saetrom,B.S.Heale,O.SnoveJr.,L.Aagaard,J.Alluin,J.J.Rossi,Distance constraintsbetweenmicroRNAtargetsitesdictateefﬁcacyandcooperativity, NucleicAcidsRes.35(2007)2333–2342.

[50]Z.Qi,S.Redding,J.Y.Lee,B.Gibb,Y.Kwon,H.Niu,W.A.Gaines,P.Sung,E.C. Greene,DNAsequencealignmentbymicrohomologysamplingduring homologousrecombination,Cell160(2015)856–869.

[51]B.Wiedenheft,S.H.Sternberg,J.A.Doudna,RNA-guidedgeneticsilencing systemsinbacteriaandarchaea,Nature482(2012)331–338.

[52]S.H.Sternberg,S.Redding,M.Jinek,E.C.Greene,J.A.Doudna,DNA interrogationbytheCRISPRRNA-guidedendonucleasecas9,Nature507 (2014)62–67.

[53]E.Semenova,M.M.Jore,K.A.Datsenko,A.Semenova,E.R.Westra,B.Wanner,J. vanderOost,S.J.Brouns,K.Severinov,Interferencebyclusteredregularly interspacedshortpalindromicrepeat(CRISPR)RNAisgovernedbyaseed sequence,Proc.Natl.Acad.Sci.U.S.A.108(2011)10098–10103.

[54]E.R.Westra,E.Semenova,K.A.Datsenko,R.N.Jackson,B.Wiedenheft,K. Severinov,S.J.Brouns,TypeI-ECRISPR-cassystemsdiscriminatetargetfrom non-targetDNAthroughbasepairing-independentPAMrecognition,PLoS Genet.9(2013)e1003742.

(10)

[55]F.Jiang,K.Zhou,L.Ma,S.Gressel,J.A.Doudna,ACas9-guideRNAcomplex preorganizedfortargetDNArecognition,Science348(2015)1477–1481.

[56]A.Graneli,C.C.Yeykal,R.B.Robertson,E.C.Greene,Long-distancelateral diffusionofhumanRad51ondouble-strandedDNA,Proc.Natl.Acad.Sci.U.S. A.103(2006)1221–1226.

[57]D.C.Swarts,M.M.Jore,E.R.Westra,Y.Zhu,J.H.Janssen,A.P.Snijders,Y.Wang, D.J.Patel,J.Berenguer,S.J.Brouns,etal.,DNA-guidedDNAinterferencebya prokaryoticargonaute,Nature507(2014)258–261.

[58]F.Gao,X.Z.Shen,F.Jiang,Y.Wu,C.Han,DNA-guidedgenomeeditingusing thenatronobacteriumgregoryiargonaute,Nat.Biotechnol.(2016)http:// www.nature.com/nbt/journal/vaop/ncurrent/full/nbt.3547.html.

[59]I.Olovnikov,K.Chan,R.Sachidanandam,D.K.Newman,A.A.Aravin,Bacterial argonautesamplesthetranscriptometoidentifyforeignDNA,Mol.Cell51 (2013)594–605.

[60]R.J.Ross,M.M.Weiner,H.Lin,PIWIproteinsandPIWI-interactingRNAsinthe soma,Nature505(2014)353–359.