ebook img

Facilitated diffusion of proteins on chromatin PDF

0.79 MB·English
Save to my drive
Quick download
Download
Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.

Preview Facilitated diffusion of proteins on chromatin

Facilitateddiffusionofproteinsonchromatin O. Be´nichou,1 C. Chevalier,1 B.Meyer,1 and R. Voituriez1 1LaboratoiredePhysiqueThe´oriquedelaMatie`reCondense´eCNRS-UPMC,4PlaceJussieu,75255ParisCedex (Dated:January27,2011) Wepresentatheoreticalmodeloffacilitateddiffusionofproteinsinthecellnucleus.Thismodel,whichtakes intoaccountthesuccessivebindingandunbindingeventsofproteinstoDNA,reliesonafractaldescriptionof thechromatinwhichhasbeenrecentlyevidencedexperimentally.Facilitateddiffusionisshownquantitativelyto befavorableforafastlocalizationofatargetlocusbyatranscriptionfactor,andeventoenabletheminimization ofthesearchtimebytuningtheaffinityofthetranscriptionfactorwithDNA.Thisstudyshowstherobustness ofthefacilitateddiffusionmechanism,invokedsofaronlyforlinearconformationsofDNA. 1 1 0 PACSnumbers: 2 n Thenowwellestablishedtheoryoffacilitateddiffusionex- modelwhichquantitativelyaddressesthesetwoquestions. a plains how DNA-binding proteins can in principle find their J target sites on DNA efficiently. This model describes search 6 trajectoriesasalternatingphasesoffreediffusioninthebulk 2 cytoplasmand1-dimensionaldiffusionalongtheDNAstrand, called sliding, which is made possible by sequence indepen- ] h dent interactions of proteins with DNA. Since the seminal c work[1],suchpathwayshavebeenevidencedexperimentally e both in vivo [2] and in vitro [3–5] thanks to single molecule m technics, and theoretical aspects have been refined [6–9], in - t particular highlighting that such strategies can minimize the a search time for a target site by a proper tuning of the pro- st tein/DNAinteraction[10–13]. FIG.1:Facilitateddiffusiononchromatin:Achromatinbindingpro- . tein (blue) searches for a target locus (yellow) on chromatin (red). t All these theoretical approaches rely on a schematic de- a Thebinding(resp.unbinding)rateisdenotedbyλ (resp.λ ). scriptionofDNAasa1-dimensionallinearchainalongwhich 2 1 m aproteincandiffuse,surroundedbyanhomogeneousmedium d- in which the protein performs regular diffusion. More re- At the theoretical level, modelling facilitated diffusion in n cently,crowdingeffectshavebeenincorporatedinthesemod- thecellnucleusraisestwoproblems: (i)first,totakeintoac- o els both for the sliding [14] and the bulk cytoplasmic phases count the switching dynamics of the protein between a state c [4], leading to more realistic descriptions of gene regulation boundtothechromatinandafreelydiffusingstateinthenu- [ kineticsinprokaryots. cleoplasm, and (ii) second to model the diffusion phase of a 2 However, such models are clearly inapplicable to eukary- proteinboundtoacomplexstructuresuchaschromatin.Point v ots, in which the DNA is packed in the cell nucleus [15] (i)hasbeenstudiedinthecontextofintermittentsearchstrate- 8 and forms a complex structure called chromatin which is far gies[21],andgeneralmethodstocalculatemeansearchtimes 5 fromasimple1-dimensionalchain. Evenifafullbottom-up forintermittenttrajectorieshavebeendeveloped.Ontheother 7 description of the in vivo DNA organization remains out of hand, the full distribution of the first-passage time (FPT) for 4 reach, theoretical ideas [16], and now growing experimental diffusion in fractals has recently been obtained in [22] and . 6 evidencesindicatethatthechromatinhasahierarchizedarchi- enables to tackle point (ii) under the assumption, backed by 0 tecturewhichdisplaysfractalpropertiesatleastoverthe100 experiments[17–19],thatthechromatinisfractal. 0 nm – 10 µm range. Indeed, textural image analysis, neutron In this letter, we gather and extend these new tools to de- 1 : scattering[17], rheologytechnics[18]andmorerecentlythe velop a theoretical model of facilitated diffusion in the cell v Hi-Cmethod[19]revealedindependentlyafractalstructureof nucleus. More precisely, we calculate analytically the mean Xi the chromatin characterized by a fractal dimension df which search time for a target for a protein which alternates diffu- wasfoundintherange2.2−3. Despitethiscomplexstructure sionphasesonthechromatinwhichisassumedtohaveafrac- r a ofthechromatin,theswitchingdynamicsofproteinsbetween talstructure,andfreediffusionphasesinthenucleoplasm(see aDNAboundstateandafreelydiffusingstate,whichcharac- Fig. 1). Underthesehypothesis, weshowquantitativelythat terizesfacilitateddiffusioninprokaryots, seemstobealsoat facilitateddiffusionineukaryotscansignificantlyspeedupthe workinthenucleus,asevidencedontheexamplesofhistons, searchprocess,andthatitenablestominimizethesearchtime high-mobility group proteins, and more generally chromatin bytuningtheaffinityoftheproteinwithDNA.Theseresults binding proteins [20]. This naturally raises the questions of arequalitativelysimilartothecaseofprokaryots,andsuggest determiningwhethertheclassicalfacilitateddiffusionmecha- that facilitated diffusion is a robust mechanism. At the theo- nismcanbeefficientalsointhecomplexnuclearenvironment, reticallevel, thisstudyyieldsasaby-productthecalculation andwhetheritcanbeusedtoregulateandoptimizegeneex- ofthedistributionoftheFPTaveragedoverthestartingpoint pression in eukaryots. This letter presents a first theoretical foraparticlediffusinginafractalstructure,whichremaineda 2 challengeinthefield[23–26]. Alternatively,onecanmakeuseoftherenewalequation[29] Search time distribution for simple diffusion in a fractal whichreadsinLaplacespace Pˆ (s) = Wˆ (s)/Wˆ (s). Us- TS TS TT medium. Wefirstconsideraproteinwhichremainsinanad- ing the symmetry relation W (t)Wstat =W (t)Wstat, the aver- ji i ij j sorbedstateanddiffusesonthechromatin,whichismodelled age over S can be taken and yields an exact expression of by a discretized domain D of volume N, and, following ex- theGFPTdistribution: Φˆ (s)=Wstat/(sWˆ (s)).Takingnext T T TT perimentalobservations[17–19],ischaracterizedbyafractal the large volume limit, the propagator Wˆ (s) can be evalu- TT dimensiondf andatypicalsizeR∝ N1/df. Wecalculatehere atedusingtheO’ShaughnessyandProcacciaformalism[30], the distribution of the search time, defined as the FPT at the whichleadsto target,averagedoverthestartingpositionoftheprotein. This (cid:16)√ (cid:17) firsttechnicalstepisnecessarytoaddresstheproblemoffacil- (cid:32) 4 (cid:33)ν Γ(1+ν) Iν A(cid:104)τT(cid:105)s aitnatiemdpdoirfftaunstiothnedoirsectuicsasledquienstthioenn.eTxhtepparroatgerianpohf;pboessiitdieosn,rit(ti)s, ΦˆT(s)= A(cid:104)τT(cid:105)s Γ(1−ν) × I−ν(cid:16)√A(cid:104)τT(cid:105)s(cid:17), (2) isassumedtoperformasymmetricnearestneighborrandom whereν=d /d ,A=2(1−ν2)/ν,andI (andlaterJ )denote walkonthechromatinwithaconstanthoppingrate(setto1). f w ν ν Besselfunctions.Introducingtherescaledvariableθ=t/(cid:104)τ (cid:105), To account for the complex organization of DNA-DNA con- T onefinallyobtainsbyinverse-Laplacetransforming(2): tactpointsthechromatincannotbedescribedasalinearchain: itiseffectivelybranchedeveniftheDNAislinear,yieldinga Q (θ) = exp(−θ) (d <d ) connectivitypotentiallylargerthan2whichtakesintoaccount T w f intersegmentaltransfer. Theresultingdynamicsischaracter- 22νν Γ(1+ν) (cid:88)∞ J (α ) = α1−2ν ν k oizfedthebymtehaenwsqalukardeimdiesnpsliaocnemdewndtewfiinthedtitmhreo:ug(cid:68)rh2(tth)e(cid:69)∝scat2l/idnwg, (1−ν2)Γ(1−ν) k=0 k J1−ν(αk) asunmdethdetonuacclteaarsmreeflmecbtrianngewwahllisc.hAbosuwnedsptrhoececehdrotmoashtionwis, tahse- × exp−2(1α2k−νν2)θ (dw >df), search time distribution is independent of these microscopic (3) details of the chromatin conformation, and is governed only by its larger scale properties which are characterized by df wheretheαk’saretherealzerosofJ−ν. andd . w WedenotebyW (t)thepropagator,i.e. theprobabilitythat ji theprotein,startingatsiteiatt=0,isatsite jattimet,and generation 6 generation 4 write Wstat for the stationary probability at site j. We will 1 theory j n make use of the pseudo-Green function of the walk defined o bainygteeHrnejesirtei=cdf(cid:82)iutn=∞n0tc(htWieojngi(ltof)b(−ta)lWwFiPsjltalTt)bd(etG,dFaenPndTo)ttehadetbLayagpiflˆva(escn)e.ttaWrragenesatfroseritmheeToref, FPT distributi0.1 a Sierpinski gasket N = 263 embedding cubes which is the FPT at site T averaged over the starting site S G N = 163 embedding cubes with weight Wstat. We thus define the probability density Φ ed 1 theory of the GFPT bSy ΦT(t) = (cid:80)Nj=1WsjtatPTj(t), where PTj is thTe rescal probabilitydensityoftheFPTatT startingfromagivensite j. b Critical percolation cluster 0.1 Generalexpressionshavebeenderivedforthefirstmoment 0 0.5 rescaled tim1e θ 1.5 2 of both the FPT [27] and the GFPT [28] (see also [23–26] forspecificexamples),andmorerecently,thehigherFPTmo- FIG.2: DistributionoftheGFPTfordiffusioninfractalmediumin mentshavebeendeterminedinthelarge-volumelimitN (cid:29)1 thecompactcase. Numericalsimulationsfordifferentsystemsizes [22]. In the case of non-compact exploration (d < d ), it N (symbols)areplottedagainstthetheoreticalpredictionofEq. 3 w f (plainlines). aSierpinskygasket(targetattheapex)andbcritical reads bond percolation cluster embedded in a 3D cubic lattice (average (cid:68)τn (cid:69)=n!(cid:104)τ (cid:105)nHTT −HTS, (1) takenoverrandomtargetsandclusterrealizations). TS T H TT where(cid:104)τ (cid:105) = H /Wstat isthemeanGFPT[28]. Usingnext This very general result, confirmed by numerical simula- that H WTstat = HTTWstTat, deduced from detailed balance, and tionsonvariousfractalsets(seefig.2and[31])showsthatthe ji i ij j (cid:68) (cid:69) GFPT distribution takes a universal form indexed by d and averaging over S, we obtain τn = n!(cid:104)τ (cid:105)n, from which it f T T dw only for any fractal structure, independently of its micro- can be deduced immediately that the GFPT distribution is a scopic details. In particular, it showsthe applicabilityof our simpleexponentialofmean(cid:104)τ (cid:105). T approachtochromatinundertheassumptionthatitisfractalat In the compact case (dw > df) ho(cid:68)weve(cid:69)r, it can be shown leastatsufficientlylargescale. Asillustrativeexamples,sim- thatduetoastrongerdependenceof τn onS [22,27],the ulations were performed on structures such as critical perco- TS averageoverS mustbetakenbeforethelargeN limit,which lationclusters,whichcapturethelargescalepropertiesofnu- makestheasymptoticformof[22]unusableforthispurpose. clearDNAorganization:theyare(i)fractalscharacterizedby 3 well defined dw and df, (ii) naturally embedded in euclidean Intermittent wona ltkh eG SMieFrpPinTs kvia graiastkioetn o wf g. ern. .t o5 λ(316 (6d sisitseos)c wiaittiho tna rfgreetq a.)t tahned a pλe2 x(rebind freq.) spaceand(iii)disordered.Thesefractalscanthereforebeseen 1.6 as minimal models of chromatin beyond the linear chain de- T1.4 P scription. F G Searchtimedistributionforfacilitateddiffusioninafractal an 1.2 e medium. Following the classical picture of facilitated diffu- d m 1 sion, we now consider that the protein can desorb from the ale cphlarsommabteinfowreitrhebraintediλn1g,taondthtehecnhrformeealtyindi(ffseueseFiing.th1e).nIuncltehois- resc00..68 λλλ222 === 000...00000115 firstapproach,weadoptameanfieldtreatmentofthephases 0.4 λ2 = 0.05 offreediffusion:weassumethatthedurationofsuchaphase λ2 = 0.1 is exponentially distributed with mean τ2 = 1/λ2, and fur- 0.2 λλ22 == 01.5 ther suppose that the protein rebinds at a position which is 0 0.1 1 10 100 1000 uniformlydistributedonthechromatin. Wedetermineinthis A τT λ1 paragraphthemeantimenecessaryfortheproteinstartingat a random position on the chromatin to reach a target locus FIG. 3: Mean search time for facilitated diffusion on a Sierpinski on the chromatin for the first time. We denote by (cid:104)T (cid:105) this gasketofgeneration5(N =366)withatargetattheapex.Numerical T mean GFPT for facilitated diffusion, and by F (t) the GFPT simulations(symbols)andtheoreticalprediction(Eq. 7,plainlines) T ofthemeanGFPT,rescaledbyitsvalueforλ = 0,isplottedasa probability density. Using tools developed in the context of √ 1 functionofx = A(cid:104)τ (cid:105)λ forvariousλ . intermittent search strategies [11, 21] , it can be shown that 0 T 1 2 the Laplace transform Fˆ is expressed in terms of the distri- T butionΦˆ oftheGFPTforsimplediffusiononthechromatin, whichisTgivenbyEq.(2): 1.6 λλ22==00..000011 NN==6115469 T1.4 λ2=0.005 N=1776  −1 FP λ2=0.05 N=620 FˆT(s)=ΦˆT(λ1+s)1− (cid:16)11−+ΦˆsT(cid:17)((cid:16)λ11++ss)(cid:17) . (4) mean G1.21 λλλ222===001.. 01 N5 = N 1N=6=922894813 λ1 λ2 aled 0.8 Assuming that diffusion on the chromatin is compact (as in resc0.6 mostexamplesoffractalsembeddedin3Dspace[32])anus- ingexpression(2)forΦˆ finallyyieldsanexplicitexpression 0.4 T oftheLaplacetransformeddistributionoftheGFPT: 0.2 0 FˆT(s)=(s+λ1)(s+λ2)Iν(xs) 0.1 1 10A τT λ1100 1000 (cid:34) Γ(1−ν) (cid:35)−1 (5) × x2νs(s+λ +λ )I (x )+λ λ I (x ) FIG. 4: Mean search time for facilitated diffusion on critical bond 4νΓ(1+ν) s 1 2 −ν s 1 2 ν s percolationclustersembeddedina3Dcubiclatticewitharandomly √ locatedtarget.Numericalsimulations(symbols)andtheoreticalpre- where xs ≡ A(cid:104)τT(cid:105)(s+λ1). Onecancheckthattheclassical diction(Eq. 7,plainlines)ofthemeanGFP√T,rescaledbyitsvalue resultoffacilitateddiffusionona1-dimensionalDNA[11,21] forλ =0,isplottedasafunctionofx = A(cid:104)τ (cid:105)λ forvariousλ 1 0 T 1 2 isrecoveredford = 1andd = 2. ThemeanGFPT(cid:104)T (cid:105)is andsystemsizeN. f w T thenreadilyobtainedbywriting (cid:104)T (cid:105)=−(cid:32)∂FˆT(cid:33) =(cid:32) 1 + 1 (cid:33)1−ΦˆT(λ1), (6) Beforecommentingonthisresult,notethatalimitdistribution T ∂s s=0 λ1 λ2 ΦˆT(λ1) GtimTeofΘth=etG/(cid:104)FTPT(cid:105)c.anDbeenoobtitnaginUed=bys(cid:104)cTon(cid:105)sidetrhiengLtahpelarceescvaalreid- T ∞ T ∞ whichfinallyyieldsthecentralresultofthispaper: able associated to Θ, and writing Gˆ (U)= Fˆ (U/(cid:104)T (cid:105) ), T T T ∞ we find using (5) in the large volume limit Gˆ (U) ∼ (1 + (cid:104)TT(cid:105)=(cid:32)λ11 + λ12(cid:33)4ΓνΓ(1(1−+νν))x02νIνI(−xν0(x)0) −1, (7) GUT)−(1Θ,)w∼hiecxhp(y−ieΘld)s. immediatelythesimpleexpTonentialform Optimalsearchtime. Wenowcommentonpreviousresults with x0 ≡ xs=0. This exact expression of the mean search and focus on the minimization of the mean search time (7) timeforfacilitateddiffusioninafractalmedium,validatedby for facilitated diffusion. First note that (cid:104)T (cid:105) ∝ N in the T ∞ numericalsimulationsonvariousexamplesoffractalswhich largevolumelimit, whichshowsimmediatelythatfacilitated mimic the large scale properties of chromatin (see Figs 3, 4 diffusion if faster than diffusion alone, which would yield a and[31])canbesimplifiedasfollowsinthelargeN limit: search time scaling as Ndw/df [28]. Actually, the search time can be minimized as a function of the desorption rate λ , as (cid:32) 1 1 (cid:33) x2νΓ(1−ν) soonastheadsorptionrateλ islargeenough. Quantitativ1ely, (cid:104)T (cid:105)∼(cid:104)T (cid:105) ≡ + 0 . (8) 2 T T ∞ λ1 λ2 4νΓ(1+ν) thefunction(cid:104)TT(cid:105)exhibitsaminimumvalueforsomeλ1 =λm1in 4 ifthevalueofthederivativeof(cid:104)T (cid:105)withrespecttoλ atλ = site even in the case of eukaryots, under the assumption that T 1 1 0isnegative. Thissetsthefollowingconditiononλ : the chromatin has a fractal organization, which seems veri- 2 fied experimentally. We add that this approach is indepen- 4ν(4−ν2) dentofthemicroscopicstructureoftheDNAandcouldalso λ (cid:62)λmin = , (9) 2 2 (cid:104)τ (cid:105)(5+2ν)(1−ν)2 be valid to some extent in the case of prokaryots, where the T DNA, even if less densely packed than in eukaryots seems whichisinpracticesatisfiedforalargeenoughchromatinvol- tohavearathercompactorganizationsignificantlydeparting ume. Underthiscondition,adirectcalculationshowsthatthe fromalinearchain. Usingtypicalexperimentalvaluesd (cid:39)3 w optimal value λm1in can be expanded in the large volume limit and df (cid:39) 2.5, one finds that the search time is minimized as: for λmin/λ (cid:39) 1/5. This suggests that the adsorption time 1 2 λmin 1−ν Γ2(ν)sin(πν)(cid:32) 2ν2 (cid:33)ν shouldbesignificantlylargerthanthetimeoffreediffusionto 1 (cid:39) − . (10) minimizethesearch, incontrastwiththeclassicalprediction λ ν π (cid:104)τ (cid:105)λ (1+ν)(1−ν)2 2 T 2 [10, 11]. This result is qualitatively compatible with experi- mentalfindings[2]onprokaryots,evenifthefractalproperties One recovers in particular for d = 1 and d = 2 the cele- f w oftheDNAstructureneedstobedeterminedinthiscase. bratedresultλmin (cid:39)λ [10,11]. 1 2 Finally,theseresultsshowthatfacilitateddiffusionisaro- bust mechanism which can speed up the search for a target [1] O.G.Berg,R.B.Winter,andP.H.vonHippel,Biochemistry [18] A.Bancaudetal.,EMBOJ28,3785(2009). 20,6929(1981). [19] E.Lieberman-Aidenetal.,Science326,289(2009). [2] J.Elf,G.-W.Li,andX.S.Xie,Science316,1191(2007). [20] T.Misteli,Science291,843(2001), [3] I.Bonnetetal.,NucleicAcidsResearch36,4118(2008). [21] O.Benichou,C.Loverdo,M.Moreau,andR.Voituriez,Physi- [4] C.Loverdoetal.,PhysicalReviewLetters102,188101(2009). calChemistryChemicalPhysics10,7059(2008). [5] B.vandenBroeketal.,PNAS105,15738(2008). [22] O.Be´nichouetal.,NatChem2,472(2010). [6] T.Hu,A.Y.Grosberg,andB.I.Shklovskii,BiophysicalJournal [23] J.J.KozakandV.Balakrishnan,PhysicalReviewE65(2002). 90,2731(2006). [24] C.P.HaynesandA.P.Roberts,PhysicalReviewE 78,041111 [7] M.SheinmanandY.Kafri,PhysicalBiology6(2009). (2008). [8] O.Benichou,Y.Kafri,M.Sheinman,andR.Voituriez,Physical [25] E.Agliari,PhysicalReviewE 77,011128(2008). ReviewLetters103,138102(2009). [26] E.AgliariandR.Burioni,PhysicalReviewE80(2009). [9] V.Dahireletal.,Phys.Rev.Lett.102,228101(2009). [27] S. Condamin et al., Nature 450, 77 (2007); O. Benichou, [10] M. Slutsky and L. A. Mirny, Biophysical Journal 87, 4021 B. Meyer, V. Tejedor, and R. Voituriez, Phys Rev Lett 101, (2004). 130601 (2008); O. Benichou and R. Voituriez, Phys Rev Lett [11] M. Coppey, O. Benichou, R. Voituriez, and M. Moreau, Bio- 100,168105(2008). phys.J.87,1640(2004). [28] V. Tejedor, O. Be´nichou, and R. Voituriez, Phys. Rev. E 80 [12] M. A. Lomholt, T. Ambjornsson, and R. Metzler, Phys. Rev. (2009). Lett.95,260603(2005). [29] S. Redner, A guide to First- Passage Processes (Cambridge [13] I. Eliazar, T. Koren, and J. Klafter, Journal of Physics: Con- UniversityPress,2001). densedMatter19(2007). [30] B.O’ShaughnessyandI.Procaccia,Phys.Rev.Lett.54(1985). [14] G.-W.Li,O.G.Berg,andJ.Elf,NatPhys5,94(2009). [31] Seesupplementaryinformationfileforotherexamples. [15] K.Bystrickyetal.,PNAS101,16495(2004). [32] D.Ben-AvrahamandS.Havlin,Diffusionandreactionsinfrac- [16] A. Grosberg, Y. Rabin, S. Havlin, and A. Neer, EPL (Euro- tals and disordered systems (Cambridge University Press, physicsLetters)23,373(1993);A.Grosberg,S.Nechaev, and 2000). E.Shakhnovich,JournaldePhysique49,2095(1988). [17] D.Lebedevetal.,CrystallographyReports53,110(2008).

See more

The list of books you might like

Most books are stored in the elastic cloud where traffic is expensive. For this reason, we have a limit on daily download.