./.
Mol.
Bid
(lQ!t;l) 227. 375-380
Characterization
of “Native” Apomyoglobin Dynamics Simulation
by Molecular
Charles L. Brooks III Department of Chemistry (larnegie Mellon ~~niwrsity Pittshuryh, PL4 1521.3. I7.S.A.
(Received IX March
1X92:
acwpted
6 Ma?y 1992)
LVe have used molecular dynamics simulation methods to study the structure and apomyoglobin in aqueous solution for a period of greater than fluctuations of “native” 05 nanosecond. This work was motivated by the recent attempts of Hughson rt al. to csharact’erize the structure and motion of both this molecule and the less compact, acid st,abilizrd I state, using methods of pulsed H/‘H exchange. The study of these systems provides new insight’s into protein folding intermediates and our simulation has yielded a detailed model for struct’ure and fluctuat,ions in apomyoglobin which complements t’he experimental studies. We find t’hat local (short-time) fluctuations agree well with fuct)uations observed for the holoprotein in aqueous solution. as well as results from thfl crystallographic H-factors. In addition. the structural features we observe for native apomyoglobin are very similar to the holoprotein, in basic agreement wit’h t,he findings of larger-scale motions, developing only over timescales in ex~ss Hugh son ef nl. By examining of a 100 picoseconds, we are able t,o identif,y conformationally “labile” and “non-labile” regions within native apomyoglobin. These regions correspond extremely well to those identified in the nuclear magnetic resonance experiments as unstable and stable “folding subdomains” in the I state of apomyoglobin. Overall we find that helices A, B, IX, G and I1 show t,he least amount of motion and helices C. I) and F move substantially over the timescales examined. The major motions. a,nd the primary difference between the holo and ape structures as we have observed them. are due to the shifting motion of helices C. I) and 17 into the vacant heme cavity. We also find that, motions at the interface of helical segments c.an be large, with one important exception being the chain segment connecting helicrs (: and H. This segment of chain interact,s with the conformationally “non-labile” helix A t’o form a relatively rigid subdomain composed of helires A. C and H. We believe that t,hestl findings provide direct support’ for t’he suggestion of Hughson et nl. that helixes A. (Z and 11 caonstitut,e a compact subdomain tha,t remains in a native-like conformation as the proteiti hegins to unfold in environment,s of decreasing pH. h’c,y/*,ords: apomyoglobin
structure; molecular dynamics; protein folding
The relationship between protein sequence and its three-dimensional structure is one of the primary unsolved problems in biology today (Anfinsen, 1973). This problem has been complicated in the past by lhe absence of stable folding int,ermediates for many single-domain globular proteins. However. a good deal of progress has been made recently in the utilization of site-directed mutagensis for the development, of mutants that form stable “folding intermediates” (Matouschck et al., 1990: Bycroft et al., 1990: Hughson & Baldwin, 1989: Hughson at al., 1991). In addiCon. the establishment of appropriate conditions to stabilize “folding intermediates” of nat ivr rccluenct3. an d the application of one- and
folding
intermediat’es:
two-dimensional n.m.r.t techniques to charact,erize these partially folded structures has led to new insights into t,he nature of protein folding int,ermediates (Hughson et al.: 1990: Matouschek et al., 1990: Raurn et al., 1989). These exciting advances are providing the motivation and fuel for‘theoretical developments in t’he modeling of protein folding. One example of such a system for which experimental studies have been carried out is apomvoglobin (Hughson et al., 1990). This prot3rin conskts t Ahhreviations used: n.m.r.. nuclear magnetic. resonance: RMSD. root-mean-square deviation: RMSF. root-mean-scluare fluctuations.
D
Figure 1. Ribbon diagram of’ the myoglobin fold. Th’) ribbon d&pm illuat~ratrs thr topology of’ the 8 helical fold of’ myoglobin, helirrs A to H arr Iat~rlrd.
of a single domain composed of eight helical segments (denot)ed A to H) and contains no disulfide bridges (a ribbon diagram of apomyogtobin is shown in Fig. 1). “Xative” apomyoglobin is produced b> removal of the p&)-porphyrin heme prosthetic group from the holopotein, which partially drst’abiliaes the tertiary fold of the gtobin. The measurable destabilization a,mounts to a decrease in thr helical content. as measured by circular dichroism from 85 ‘!() in the holoprotein t.0 about 550z’, in native apomyogtobin (Griko et nl., 1988). In addition, reduc+on of t’he pH from neutral values to pH 4 to 1.5 produces a stable intermediate with helical c.ont.ent of about 35°~o (Griko et nl.. 1988). This state is trrmrd the I state. Further reduction of pH t)o near pH 2 yields a structurt, that appears as random-coil. The ability to creat,r stable. part)ially folded st’ates of nlyoglobin. bot’h native a~pomyoglobin and thr, I stat?. has prompted recent interest in the st,udy of this syst,rm by biochemical and spet8troscBopica methods. I’sing the t’echniqurs of pulsed H/‘H csc-hangr together wit’h oneand t Mo-ditrlr,nsional n.m.r. experiments. t’he Haldwin laboratorv has begun t,o providr fairly det’ailed structural information regarding the nat,urr of these states (Hughson et al.. 1990). This st,ruc%ural information suggests the presence of c*onformationalty “labile” and “nonlabile” secondary structural elements which may be the precursors of late and early forming structural intermediates in t)he folding pathway of myogtobin. From an analysis of their findings on the protect’ion of exchangeable backbone amide protons. Hughson et al. (1990) tzont4ude that the native apomyoglobin st8ructurr is very similar to that of the holoprotrin. However, they also note that their findings do not) fully explain the difference in helical cont’ent.
Thtb initial findings of’ Hu&on rt trl. (I!)!)(I) havt, rnotivat~~d our work OII this s;?Aeni. In i)art icular. wc havts bcpn to invest ip1t.r‘ ii tiumbt~r of ‘pst ions rt~gar’lin~ t ht. nat II re of sU’*ll St rll(‘tUlYil illttbr mtxdiat’rs ill aqueous solution using nit~iho~ls of’ niole’~ular ‘iynami’*s. Our initial invt5tiptions. tiesc~ribt~d in this ‘~orrirrii~~~i’~i~tiori. havta fi)causst4 otl the quest ions of whether olle (*an itient,if:\, c’otlfi)rrrtiltionall>labile and rioti-lahiltt fielding qions iI1 IlittiVCt ;t~)ot”!‘o~ilot,ili under cx)tiditions sirtlilitr t 0 those studi by Hughsoll rt r/l. (1990). Our i)ritnar,v objrctivt3 art itinlC3l at providing tl’Jtilil’Yl ilrll( tural modr>ls for protein ii)lding intc~rm~tliatt~s t Ililt arc t,estabtt~ against ‘~urrt~n1 ‘~sprirriettt al rt231115 and ot tr’~r t hrloret itxl nio’l’als. 1‘11~ simulation \v“ ha\-tl c.arritld out (‘onsihls of’ over 500 ps of‘ rnolf33lliLr clynamics fC)r apo”~yoglobin (grtitc~te’l from t hr, sl ru’.t III’~ of’ spwrl \vhalt~ myogtobin (Phillips, 1980) I)>- rt~movai of t trcs I)ros thctic group follo\vtatl [)J- ‘~nrrg~~ minitnizaI ion 1 o reli’,v’x t)ild c.olltatats) ill 4299 \vat.rr rrlott~c*ult3. ‘I’ll amounl of’ solvrrll inc4irdt4 itI out’ sl lidit ih of’ sufficit\nt siztb to provi’lo a (i 10 X:4 (1 .;i = 0.1 nm) watrr tavrr around i~~~O~ll~o~lot~ir1. 12’ca rlsrcl (‘01Ivt~I1 tional m’olecautar dynamic:s mt~thotls as irnf)tt,ttic,rltt,ti in t~ll’~ (‘HARMIRI tn;~~crornott~c~~~t~~r ciynarnics program (Brooks rt ol., 1983: Mrr1 z et (11.. I!)!)1 ) \vit h i t,iniestep of 04)02 ps. Thr pAf+iti \vas rt~prt~sr~tllt~tl with the (‘HAKMJI vrrsion 19 polar hy(lrogt~rl f)a~‘~~meter stlt (on1y hydropns that (YLI~ fijrtn h?-ctrog;t~rl bonded interac*tions arr included thxf,ticit ty. tlytlrtr grns ittt&C’htYi to ‘.arbon ill’? itlc4udctl inii)tic,it ty: Brooks c,t (I/.. 1984) alltl t hch a,atrr \vits rt~~)rc~sc~tltc~cl by the ‘1’11’3I’ rnodt~l of Jorgc~trst~n it r/l. (I!)x:lj. ‘I’ht~ water and all tivdropn LttOln tl6’klV~ iltOr)l I)Olltlh \vere trrattatl ils Lx4 rising t hr, SH:\I\Ij: iLlgot,il htrl (Kyckaprt et cl/.. 1977: Slc*rfz rt c/I.. I!)!)1 ). I’t~riotiic* boundary cnonditions \\t~rc~~~~nif~to~c~tl withilr 3 c,ul)ic, simulation \otIlmt~ of 5.5 A otl il si&. NotI-l)o~rdt~cl neighbor lists \Vt”re !g(~lWlXttVl to ‘~xptliic~ (‘itlt’llt~if ioti of thr elect rostatica and vi111 tlrbr L$‘?littS inttLrac4 ioii 0
__-.
Cvmmunicntions
377
Figure 2. (” trace of my&bin cqstal structure and apomyoglobin tlynan1ic.s avrragrd structurr supc‘rimposrd. A stereo virw of thr crystallographic co-ordinates for myoglobin (lighter line) and average structurr from 420 lx of molecular dynamics of apomyoglobin (heavier line) illustrate thta overall topological similarity of thr 2 structurrs as well as regions of thr aJ)o-structure which move away from their starting X-ray values. truncation scheme. which shifts the forces to zero at thv cutoffdistancbe. This truncation scheme has I)ern shown to yield very good results for t’hr structure of pure water (f3rooks rt al., 198.5). Finally. the densit? of water away from the protein a-as near 1 g/cc8 and the temperature wa,s maintained near 318K by reassigning velocities from a Hoftzmann distribution characteristic. of this temperature if the temperature (measured as the mclan kinetica energy over a minimum of %OO steps of dynamics) drifted outside a window of f 10 ‘K. The period of analysis upon which we will focus in this c~ornmunication is the first 0.5 nst. Tn particular. wc consider the period of 0 to 80 ps as rquilihration and 80 to 500 ps as dynamics for a,nafysis. The initial stages of dynamics (0 to 8Ops) are characterized by motion of t,he structure from its start.ing position. i.e. the crystallographic coordinates for the globin from the sperm whale myglobin structure deposited in t,hr Krookharrn Protein IIatabank (Rernstein rt (I/.. 1977) by Phillips (Phillips, 1980). The overall root-mearrs;ciua”t~-d~?viation (RMKD) of the structure from the crystal st’ructurr rrached a plateau at 3.2 A RMSI> around 6Ops into the equilibration phase. At t’his point the major differences between the starting crystallographic structure and the molecular dyn”mics struct,ure had developed and wp considered cyuifibration caompletr. Cl’e simulated the system for an additional 20 ps to be certain this was the case hefore we began the analysis phase. From t’he simulation period X0 to 500 ps we computed thta average structure and fluctuations about t,his si.ruct~urr. Since the protein molecule rot at,ed by - 1.5” from its init,ial orientat,ion, it was necessary to reorient the tnolrcuf~~ hack to a common reference frame t)o c0nstruc.t reasonable average structural quantit)ies. \1’c did this 1)~ csompiiting suhaveragfss for structure and fiuc%uc t Our simulation to date has been carried out in of 900 ps. Drtailetl analysis of this simulat,ion under way and will be presmt~ed in a subsquent public2ltion.
escfw
is
tions during 10 1)s “windows” (a, tirntk period over which rot’ation was minimal) followrtl 1)~ overlaying ctach of the 4% subaverage structures and averaging to yield the final structure (for an average over a total of 410 1)s). The C‘” t.racr of this structure is compared with the st,arting crystal st,ruc%ure in Figure %. From this Figure we notr t,hat t.he topology of the fold is similar for both our molec~ufar dynamics strucbturfa and t,he S-ray structure. IJifferences in the positioning, and degree. of secondary strucat ure present> occur predominantly for helicrs H to F. Notion in thih region occurs primarily as a result of thta “ca\-ing 111 ” of the vacant hrmr cavity. ,Itlditionaf differences that occur arc’ t,ltr fraying of the (‘- and X t,erminal segments of helicrs A and H. ‘I’llis structural picture is very similar to the holoprotein and fully consistent wit,11 t’he patterns of 1)roton rxcaha.ngr protection and t’hc structural mod?1 of Hughson c~f 01. (lU90) for native apomyoplobin. Sext we examine fluctuat’ions of segments within t flta structure. Regions of thr structure from molecular d\-natnics which show movrment with resptBc+ to the cqstallographic positions might also be anticipat)ed to be regions of high flexibility. For this reason wt’ separat)e the locbal (short-time) flurtuations from larger-scale (long-time) fluctuations. This was achitlved by computing the long-time fluctuations as the average mean-square flllctuatJions of t ht> ten picbosrcond average structures about the -IN ps averagt’ structure. The rcbsnlts of t,hese findings are shown in Figure 3. whrrr the residueavt~ra~ged t,oot-rnran-scluare fluctuations (RMSF, comput,rd ilsing all atoms ass0ciat.A with a given rtbsidue) are plotted 7wsus residue number. Furthrrmore. to “smoot 11” the transition between regions of high and low mobility. wta have displayed thtlse averages as a running window a\-erage over rAdues on either side of the c*rntral residue. i.e. a n-indow of width three. As is evident from comparison of the continuous line (local fluctuations) and t 1~ broken line (larger-scale fluctuations). focal Huct nat,ions yield residue-averaged RMSF values of approximatf~l>~ 0.5 to 0.75 A. wfif~rt~as the larger-
2.0
A
,B 20
C 40
- D,
E 60
F I 80 Residue number
H
G loo
120
140
Figure 3. Root,-mean quart’ fluct,uations of residues in m)oglobin. The RMSF (in ;i) f or apornyoglot~in ar'fs plott,f~cl ti)r local (short-time) motions as averages over 10 ps “\z indo\vs” from the 420 ps molecular dynamics simulation in t,hr c*olltinuous line. the broken line shows larger-soalr (long-timr) average Huvtuations c,onstructrd f’rorll th SHIW simulation. (‘omparisons are made with the R,MSF from both cryst,allographic. f&factors (drnotrd with synl)ol 0) imd molrvular dynamics simulation (denoted with symbol + ) for the nativr holo protein. The RMSF are plott~t~d wrsu.~ residue number using a running residue average over 3 rrsiclurs to smooth t,he resulting dat,a. Thr loczal fluctuations agrtat \brll with both X-ray ant1 molecular dynamics results for thr holo protjein. I,arger-scale fiuctuations in h&crs (‘. I) antI F suggvst c~onfi,rmational “lahilit,y” of native apomyoglohin in t,hrse regions of thtb struc%ur’e.
scale fluctuat,ions ark a factor of two to three t,irmBs grrhater. In addition, we find that) the local fluct,uaCons are “t,ypical” of RMSF values seen in the holoprotein. This is illustrat’ed by caomparison with bot)h the RMSF values computed from the crystallographic /I-fa&ors (Phillips. 1980) and those from a previous simulat,ion of holomyoglobin in aqueous solution (Findsen it nl., 1991). Furthermore. these motions are representative of motions occurring in unperturbed protein structures at) room-temperaturr over periods of less than 100 t.o 200 ps (Brooks f,‘f al.. 1988; McCammon & Harvey, ISXi). The larger-scale fluctuations displayed in Figure 3 provide deeper insight into the nature of “segmental” motions occurring over longer tjime scales in apomyoglobin and arp more representative of the motions expected to occur in solut,iotl as st,udied by n.m.r. (Hughson rt /A.. 1990: C~CWCO& I,ec*omtIr. 1990). It’ was necessary to extend our simulations beyond the few hundred picosecaonds usually usrd in studies of this type t’o observe such mot,ions under conditions t’hat are similar t.o those used in the laborat~ory. In this Figure we see regions of high mobility. as indicated by large RMSF values and regions of much lower mobilit’y. \;Z’e would suggest that, the regic~ns of the structurr with RMSF values less than 1.2 A (which is chosen rather arbtrarily simply to guide our discussions) would repr+ sent conformationally non-labile units while the other we will trrrn conformationally labile. LVit’h
this qualitative classification we find that helic*t+ A. t3, and E to H have regions of non-lability ovtlr these Gmescales whereas helixes (‘ and 1) are Iabilc~. Purt~hermore, helices K. E and (: are thr most M)I~labilr. Helices A and H have regions away from t tic helical core which undrrgo motions but appear as predominantly non-labile hrlical segments. Mot ion in helix F increases steadily from the S-trrminnl rnd into the F(: corner. It is this section of tht, chain t,hat rnovrs into the vacant hemr (*avita?. The tyqion including heticrs (’ and I) shows larger-sc*alc tnotiorrs occurring over’ the timesc.alrs of 300 to 100 ps studied here. Wr notje that’ under t ht relativc4y mild conditions studied here (pH near 7 and tPrnprraturc> around 31dK) these motions represent) c~quilihrium fluctuations. How-ever. we bcllieve these motionh arc’ the indicators of structural regions which an’ “labile” towards unfolding. which would takr place at higher tc~mperaturt~s or a,t lower pH. These motions arc also illustrated by t ht, suprsrposition of the ten picosrc*ond avckrage slruc*t ures shown as in a stereo view of t)heir (‘a traccls in Figurrs 1. We havr chosen structurrs spa~d 10 pic,osec*onds apart, for clarity of t.hr IQgure. From this Figure. regions of t.he molec~ule that arp relativrly notelabile. i.e. those that overlay wf,II in thr super’posiGon, and t,hosr t)hat are labile are pasil). observed. Labile regions such as the, (!-terminal rnd of the R helix t,hrough the N-terminal end of the I< hctlix are c*lrarl>, obsrr\-c>cl in thr Figure with t hta
Figure 4. SuJwrposition of (I” tracrs from average structurrs throughout thp 120 ps molecular dynamiw simulation of is displayed with sampling of one structure apc,myogiot)in. A st,rreo view of structures averaged over 10 ps “windows” r\-rr~’ 40 ps. Thrl large range of longer timescale motions. wJwiall?; for the (1. 1) helical region and FG wrner. is illustrated 1)~ the “scat&r” of structural elements in these regions. In addition. thr stability of thr (: to H wnnrctivr IooJ~. as it J&ks against the A hrllix is noted.
largta range of’ motion present, in thta (‘-1) helical region being outstanding. Similar observations can he made regarding the FG corner with the adjacent. helical termini also displaying motion. Tt is intrresting to note that the non-helical region connecting hrfix G with H shows mohifitSy less than the (‘-I) helical region. This may occur because it packs against the relat,ively non-labile A helix. Toget’her. heficc>s A, C; and H form a subdomain showing little motion in our simulation. This suggests stability of this unit. whicah is in agreement with the hypothesis of Hughson rt ~1. (1990). While it is difficult to make direct comparison with the results from recent n.m.r. experiments (Hughson it N/.. 1990). it is rxt)remefy encouraging that our resuftjs from molecular dynamics calcufat)ions are in good acacbord with the qualitative picture of slruc%ut*e and flexibility in apomyoglobin developed from these experiments. Il’e have identified dXrrenc*rs in t’he motional characterist’ics for regions of thr protein. which we term c-onformationafly labile and non-labile. These regions correspond well with the proton exchange protection factors ot)swved by t1.m.r. (Hughson et al., 1990). a,nd lend support to the suggestion of Hughson rt nl. (1990) that trclliczes A, (: and H might) form a stable folding unit in the I state of apomyoglobin. Furthermore. we fitId tha.t substantial motion owurs in regions adjacent to the home pocket, hefices (‘. I> and F. as well as thr segments directly preceding and following these regions. as a result, of thr absence ot t ht. prosthet ic. group and exposure to solvent. Motions wt’ observed suggest t,hat multiJ)le corlformations ma: contribute to XOE couplings in t flese regions of the sequence and this may compli(~1 e full determination of the solution structure of native apom>-ogfobin hy n.m.r. methods ((I~cc*o & I,ewmte. I 990). fIrsf& our observation t)hat signiticsant motion occurs over timescales in the 200 to 400 ps range. the overall structure atld focal fluctuations we observe for the native state of af)orr~?‘o~fof)il~ are yuitrb similar to t.flose for its hofo
c~ounterpart. This t)oo is in basic agreement with the findings from experimental st,udirs on t.his &em. A final point worth mentioning is the timrscafe required t’o “see” larger-scale mot’ions under condiCons of t,emperahure and pH similar to t,hose being explored in the current n.m.r. experiments. It was necessary to rxbend the molecular dynamic-s simulations beyond the - 100 to %o() [JS timescafe used typicsally for sofvat’ed protein systems of this size. all d we are continuing our analysis of the sirnufations described in this communiration to periods greater than 900 ps. This necessity is a direcat) c‘onsrquence of our desire to observe t,he system near experimental conditions. It is probable t,hat increasing the temperature (Tirado-ISives & .Jorgensen. 1991: IXapua d al., 1990: Fan rt al., 1991: Soman fjt al., 1991; Dagpert &, Levitt. 1992) 01 modifying the nat,ure of the inter-at’ornic* interactions (for example, by making all atoms “see each other” through purely repulsive rlet*trost,atic int’eractions (Sehnften, personal comrnunicaation)) may !,ield a similar pic.t>urr of c~onformat,ionafly labile /‘prsTls non-labile groups. However. since one of our J,rimary objectives is to provide det’a,iled structural models for systems currently under experimental sc*tutiny, we feel that these met’hods are not viable. Instead. we have focussed upon ryuifibrium-based met,hods that permit us to develop nru-native. noncotnJ)act stattbs in aqueous solution. and we will f)ursue the equilibrium studies of i hew novrf stat)es of’ J)art,iafly folded protein in future f)ubficxtions. ‘l’hca author \vc~nltl likta t,o acknowlrdpr thr following agrhncaies fhr Jjartial financial support of’ this work: the Sational TnsCtutes of Health (GM 375.54: 1’4 I RJtMOO9). the Kational Science Foundation (.AL\s(‘X!W?XL)6) and A. J’. Sloan Foundation for fellowship suJ)port. SJw+al t,hanks is owed to I)r *J. Mertz of (:raJ. Resrarc*h. Inc.. for his invaluahlr role in helping with somts of t hr c,alculations rrywrted herein. 1%‘~ are gratrful to the J’ittsburgh SrlJ)rrc~omJ)utiriy (‘rntrl and (?a!, Itcwarc~ti. Inc. for sul)pJying the wmputer rwourws ~~ecwsary to c.arr\- out thrse c~:rl(~ulat~ions. Thr, author woultl also likca tc, thank I)r
3x0
( ‘.
----__-
.\I. 1’clttit.t for elf’ ttolortt?;oplol)in
sharing
his RMSF data solution (Findstw
irl
t’ron~
I,.
I,‘rooks
I I I
_-
hinrulatilltl I!J!ll).
thy
rt ~1..
Iliturn.
.I.
structnrc~
c1vnarnic.s
1'58. Findsrn.
References IZllrn.
M.
I’.
Xinrrrlation
B.
Tildrslr~.
of
Liqrcitls.
Haurn.
(‘. protrin
H. (1973). chain.
I). .J. Oxford
.I ., ( 1989).
I’rinciples
(1989). I’nivctrsity
I)ohson. (‘. (‘hara~terizttt,ion
SMR methods: all)hcc-la~talhumin. Rvrnstrin. Mc~rr.
F. E.
that
181.
A’cirnce.
%I..
the
I’rrss.
A. & folded
C’.. Koetzle. F., Krirr. M.
28.
‘I’. I)..
(iriko.
folding
Hartley. protein
(‘. l)J
guinea
pig
globule state of 7-l
:1.
F.. \Villiamh. Rodgers. .I.
R..
(:. ‘I’. Kinnard.
I
Mol.
Bid
(lQ!t;l) 227. 375-380
Characterization
of “Native” Apomyoglobin Dynamics Simulation
by Molecular
Charles L. Brooks III Department of Chemistry (larnegie Mellon ~~niwrsity Pittshuryh, PL4 1521.3. I7.S.A.
(Received IX March
1X92:
acwpted
6 Ma?y 1992)
LVe have used molecular dynamics simulation methods to study the structure and apomyoglobin in aqueous solution for a period of greater than fluctuations of “native” 05 nanosecond. This work was motivated by the recent attempts of Hughson rt al. to csharact’erize the structure and motion of both this molecule and the less compact, acid st,abilizrd I state, using methods of pulsed H/‘H exchange. The study of these systems provides new insight’s into protein folding intermediates and our simulation has yielded a detailed model for struct’ure and fluctuat,ions in apomyoglobin which complements t’he experimental studies. We find t’hat local (short-time) fluctuations agree well with fuct)uations observed for the holoprotein in aqueous solution. as well as results from thfl crystallographic H-factors. In addition. the structural features we observe for native apomyoglobin are very similar to the holoprotein, in basic agreement wit’h t,he findings of larger-scale motions, developing only over timescales in ex~ss Hugh son ef nl. By examining of a 100 picoseconds, we are able t,o identif,y conformationally “labile” and “non-labile” regions within native apomyoglobin. These regions correspond extremely well to those identified in the nuclear magnetic resonance experiments as unstable and stable “folding subdomains” in the I state of apomyoglobin. Overall we find that helices A, B, IX, G and I1 show t,he least amount of motion and helices C. I) and F move substantially over the timescales examined. The major motions. a,nd the primary difference between the holo and ape structures as we have observed them. are due to the shifting motion of helices C. I) and 17 into the vacant heme cavity. We also find that, motions at the interface of helical segments c.an be large, with one important exception being the chain segment connecting helicrs (: and H. This segment of chain interact,s with the conformationally “non-labile” helix A t’o form a relatively rigid subdomain composed of helires A. C and H. We believe that t,hestl findings provide direct support’ for t’he suggestion of Hughson et nl. that helixes A. (Z and 11 caonstitut,e a compact subdomain tha,t remains in a native-like conformation as the proteiti hegins to unfold in environment,s of decreasing pH. h’c,y/*,ords: apomyoglobin
structure; molecular dynamics; protein folding
The relationship between protein sequence and its three-dimensional structure is one of the primary unsolved problems in biology today (Anfinsen, 1973). This problem has been complicated in the past by lhe absence of stable folding int,ermediates for many single-domain globular proteins. However. a good deal of progress has been made recently in the utilization of site-directed mutagensis for the development, of mutants that form stable “folding intermediates” (Matouschck et al., 1990: Bycroft et al., 1990: Hughson & Baldwin, 1989: Hughson at al., 1991). In addiCon. the establishment of appropriate conditions to stabilize “folding intermediates” of nat ivr rccluenct3. an d the application of one- and
folding
intermediat’es:
two-dimensional n.m.r.t techniques to charact,erize these partially folded structures has led to new insights into t,he nature of protein folding int,ermediates (Hughson et al.: 1990: Matouschek et al., 1990: Raurn et al., 1989). These exciting advances are providing the motivation and fuel for‘theoretical developments in t’he modeling of protein folding. One example of such a system for which experimental studies have been carried out is apomvoglobin (Hughson et al., 1990). This prot3rin conskts t Ahhreviations used: n.m.r.. nuclear magnetic. resonance: RMSD. root-mean-square deviation: RMSF. root-mean-scluare fluctuations.
D
Figure 1. Ribbon diagram of’ the myoglobin fold. Th’) ribbon d&pm illuat~ratrs thr topology of’ the 8 helical fold of’ myoglobin, helirrs A to H arr Iat~rlrd.
of a single domain composed of eight helical segments (denot)ed A to H) and contains no disulfide bridges (a ribbon diagram of apomyogtobin is shown in Fig. 1). “Xative” apomyoglobin is produced b> removal of the p&)-porphyrin heme prosthetic group from the holopotein, which partially drst’abiliaes the tertiary fold of the gtobin. The measurable destabilization a,mounts to a decrease in thr helical content. as measured by circular dichroism from 85 ‘!() in the holoprotein t.0 about 550z’, in native apomyogtobin (Griko et nl., 1988). In addition, reduc+on of t’he pH from neutral values to pH 4 to 1.5 produces a stable intermediate with helical c.ont.ent of about 35°~o (Griko et nl.. 1988). This state is trrmrd the I state. Further reduction of pH t)o near pH 2 yields a structurt, that appears as random-coil. The ability to creat,r stable. part)ially folded st’ates of nlyoglobin. bot’h native a~pomyoglobin and thr, I stat?. has prompted recent interest in the st,udy of this syst,rm by biochemical and spet8troscBopica methods. I’sing the t’echniqurs of pulsed H/‘H csc-hangr together wit’h oneand t Mo-ditrlr,nsional n.m.r. experiments. t’he Haldwin laboratorv has begun t,o providr fairly det’ailed structural information regarding the nat,urr of these states (Hughson et al.. 1990). This st,ruc%ural information suggests the presence of c*onformationalty “labile” and “nonlabile” secondary structural elements which may be the precursors of late and early forming structural intermediates in t)he folding pathway of myogtobin. From an analysis of their findings on the protect’ion of exchangeable backbone amide protons. Hughson et al. (1990) tzont4ude that the native apomyoglobin st8ructurr is very similar to that of the holoprotrin. However, they also note that their findings do not) fully explain the difference in helical cont’ent.
Thtb initial findings of’ Hu&on rt trl. (I!)!)(I) havt, rnotivat~~d our work OII this s;?Aeni. In i)art icular. wc havts bcpn to invest ip1t.r‘ ii tiumbt~r of ‘pst ions rt~gar’lin~ t ht. nat II re of sU’*ll St rll(‘tUlYil illttbr mtxdiat’rs ill aqueous solution using nit~iho~ls of’ niole’~ular ‘iynami’*s. Our initial invt5tiptions. tiesc~ribt~d in this ‘~orrirrii~~~i’~i~tiori. havta fi)causst4 otl the quest ions of whether olle (*an itient,if:\, c’otlfi)rrrtiltionall>labile and rioti-lahiltt fielding qions iI1 IlittiVCt ;t~)ot”!‘o~ilot,ili under cx)tiditions sirtlilitr t 0 those studi by Hughsoll rt r/l. (1990). Our i)ritnar,v objrctivt3 art itinlC3l at providing tl’Jtilil’Yl ilrll( tural modr>ls for protein ii)lding intc~rm~tliatt~s t Ililt arc t,estabtt~ against ‘~urrt~n1 ‘~sprirriettt al rt231115 and ot tr’~r t hrloret itxl nio’l’als. 1‘11~ simulation \v“ ha\-tl c.arritld out (‘onsihls of’ over 500 ps of‘ rnolf33lliLr clynamics fC)r apo”~yoglobin (grtitc~te’l from t hr, sl ru’.t III’~ of’ spwrl \vhalt~ myogtobin (Phillips, 1980) I)>- rt~movai of t trcs I)ros thctic group follo\vtatl [)J- ‘~nrrg~~ minitnizaI ion 1 o reli’,v’x t)ild c.olltatats) ill 4299 \vat.rr rrlott~c*ult3. ‘I’ll amounl of’ solvrrll inc4irdt4 itI out’ sl lidit ih of’ sufficit\nt siztb to provi’lo a (i 10 X:4 (1 .;i = 0.1 nm) watrr tavrr around i~~~O~ll~o~lot~ir1. 12’ca rlsrcl (‘01Ivt~I1 tional m’olecautar dynamic:s mt~thotls as irnf)tt,ttic,rltt,ti in t~ll’~ (‘HARMIRI tn;~~crornott~c~~~t~~r ciynarnics program (Brooks rt ol., 1983: Mrr1 z et (11.. I!)!)1 ) \vit h i t,iniestep of 04)02 ps. Thr pAf+iti \vas rt~prt~sr~tllt~tl with the (‘HAKMJI vrrsion 19 polar hy(lrogt~rl f)a~‘~~meter stlt (on1y hydropns that (YLI~ fijrtn h?-ctrog;t~rl bonded interac*tions arr included thxf,ticit ty. tlytlrtr grns ittt&C’htYi to ‘.arbon ill’? itlc4udctl inii)tic,it ty: Brooks c,t (I/.. 1984) alltl t hch a,atrr \vits rt~~)rc~sc~tltc~cl by the ‘1’11’3I’ rnodt~l of Jorgc~trst~n it r/l. (I!)x:lj. ‘I’ht~ water and all tivdropn LttOln tl6’klV~ iltOr)l I)Olltlh \vere trrattatl ils Lx4 rising t hr, SH:\I\Ij: iLlgot,il htrl (Kyckaprt et cl/.. 1977: Slc*rfz rt c/I.. I!)!)1 ). I’t~riotiic* boundary cnonditions \\t~rc~~~~nif~to~c~tl withilr 3 c,ul)ic, simulation \otIlmt~ of 5.5 A otl il si&. NotI-l)o~rdt~cl neighbor lists \Vt”re !g(~lWlXttVl to ‘~xptliic~ (‘itlt’llt~if ioti of thr elect rostatica and vi111 tlrbr L$‘?littS inttLrac4 ioii 0
__-.
Cvmmunicntions
377
Figure 2. (” trace of my&bin cqstal structure and apomyoglobin tlynan1ic.s avrragrd structurr supc‘rimposrd. A stereo virw of thr crystallographic co-ordinates for myoglobin (lighter line) and average structurr from 420 lx of molecular dynamics of apomyoglobin (heavier line) illustrate thta overall topological similarity of thr 2 structurrs as well as regions of thr aJ)o-structure which move away from their starting X-ray values. truncation scheme. which shifts the forces to zero at thv cutoffdistancbe. This truncation scheme has I)ern shown to yield very good results for t’hr structure of pure water (f3rooks rt al., 198.5). Finally. the densit? of water away from the protein a-as near 1 g/cc8 and the temperature wa,s maintained near 318K by reassigning velocities from a Hoftzmann distribution characteristic. of this temperature if the temperature (measured as the mclan kinetica energy over a minimum of %OO steps of dynamics) drifted outside a window of f 10 ‘K. The period of analysis upon which we will focus in this c~ornmunication is the first 0.5 nst. Tn particular. wc consider the period of 0 to 80 ps as rquilihration and 80 to 500 ps as dynamics for a,nafysis. The initial stages of dynamics (0 to 8Ops) are characterized by motion of t,he structure from its start.ing position. i.e. the crystallographic coordinates for the globin from the sperm whale myglobin structure deposited in t,hr Krookharrn Protein IIatabank (Rernstein rt (I/.. 1977) by Phillips (Phillips, 1980). The overall root-mearrs;ciua”t~-d~?viation (RMKD) of the structure from the crystal st’ructurr rrached a plateau at 3.2 A RMSI> around 6Ops into the equilibration phase. At t’his point the major differences between the starting crystallographic structure and the molecular dyn”mics struct,ure had developed and wp considered cyuifibration caompletr. Cl’e simulated the system for an additional 20 ps to be certain this was the case hefore we began the analysis phase. From t’he simulation period X0 to 500 ps we computed thta average structure and fluctuations about t,his si.ruct~urr. Since the protein molecule rot at,ed by - 1.5” from its init,ial orientat,ion, it was necessary to reorient the tnolrcuf~~ hack to a common reference frame t)o c0nstruc.t reasonable average structural quantit)ies. \1’c did this 1)~ csompiiting suhaveragfss for structure and fiuc%uc t Our simulation to date has been carried out in of 900 ps. Drtailetl analysis of this simulat,ion under way and will be presmt~ed in a subsquent public2ltion.
escfw
is
tions during 10 1)s “windows” (a, tirntk period over which rot’ation was minimal) followrtl 1)~ overlaying ctach of the 4% subaverage structures and averaging to yield the final structure (for an average over a total of 410 1)s). The C‘” t.racr of this structure is compared with the st,arting crystal st,ruc%ure in Figure %. From this Figure we notr t,hat t.he topology of the fold is similar for both our molec~ufar dynamics strucbturfa and t,he S-ray structure. IJifferences in the positioning, and degree. of secondary strucat ure present> occur predominantly for helicrs H to F. Notion in thih region occurs primarily as a result of thta “ca\-ing 111 ” of the vacant hrmr cavity. ,Itlditionaf differences that occur arc’ t,ltr fraying of the (‘- and X t,erminal segments of helicrs A and H. ‘I’llis structural picture is very similar to the holoprotein and fully consistent wit,11 t’he patterns of 1)roton rxcaha.ngr protection and t’hc structural mod?1 of Hughson c~f 01. (lU90) for native apomyoplobin. Sext we examine fluctuat’ions of segments within t flta structure. Regions of thr structure from molecular d\-natnics which show movrment with resptBc+ to the cqstallographic positions might also be anticipat)ed to be regions of high flexibility. For this reason wt’ separat)e the locbal (short-time) flurtuations from larger-scale (long-time) fluctuations. This was achitlved by computing the long-time fluctuations as the average mean-square flllctuatJions of t ht> ten picbosrcond average structures about the -IN ps averagt’ structure. The rcbsnlts of t,hese findings are shown in Figure 3. whrrr the residueavt~ra~ged t,oot-rnran-scluare fluctuations (RMSF, comput,rd ilsing all atoms ass0ciat.A with a given rtbsidue) are plotted 7wsus residue number. Furthrrmore. to “smoot 11” the transition between regions of high and low mobility. wta have displayed thtlse averages as a running window a\-erage over rAdues on either side of the c*rntral residue. i.e. a n-indow of width three. As is evident from comparison of the continuous line (local fluctuations) and t 1~ broken line (larger-scale fluctuations). focal Huct nat,ions yield residue-averaged RMSF values of approximatf~l>~ 0.5 to 0.75 A. wfif~rt~as the larger-
2.0
A
,B 20
C 40
- D,
E 60
F I 80 Residue number
H
G loo
120
140
Figure 3. Root,-mean quart’ fluct,uations of residues in m)oglobin. The RMSF (in ;i) f or apornyoglot~in ar'fs plott,f~cl ti)r local (short-time) motions as averages over 10 ps “\z indo\vs” from the 420 ps molecular dynamics simulation in t,hr c*olltinuous line. the broken line shows larger-soalr (long-timr) average Huvtuations c,onstructrd f’rorll th SHIW simulation. (‘omparisons are made with the R,MSF from both cryst,allographic. f&factors (drnotrd with synl)ol 0) imd molrvular dynamics simulation (denoted with symbol + ) for the nativr holo protein. The RMSF are plott~t~d wrsu.~ residue number using a running residue average over 3 rrsiclurs to smooth t,he resulting dat,a. Thr loczal fluctuations agrtat \brll with both X-ray ant1 molecular dynamics results for thr holo protjein. I,arger-scale fiuctuations in h&crs (‘. I) antI F suggvst c~onfi,rmational “lahilit,y” of native apomyoglohin in t,hrse regions of thtb struc%ur’e.
scale fluctuat,ions ark a factor of two to three t,irmBs grrhater. In addition, we find that) the local fluct,uaCons are “t,ypical” of RMSF values seen in the holoprotein. This is illustrat’ed by caomparison with bot)h the RMSF values computed from the crystallographic /I-fa&ors (Phillips. 1980) and those from a previous simulat,ion of holomyoglobin in aqueous solution (Findsen it nl., 1991). Furthermore. these motions are representative of motions occurring in unperturbed protein structures at) room-temperaturr over periods of less than 100 t.o 200 ps (Brooks f,‘f al.. 1988; McCammon & Harvey, ISXi). The larger-scale fluctuations displayed in Figure 3 provide deeper insight into the nature of “segmental” motions occurring over longer tjime scales in apomyoglobin and arp more representative of the motions expected to occur in solut,iotl as st,udied by n.m.r. (Hughson rt /A.. 1990: C~CWCO& I,ec*omtIr. 1990). It’ was necessary to extend our simulations beyond the few hundred picosecaonds usually usrd in studies of this type t’o observe such mot,ions under conditions t’hat are similar t.o those used in the laborat~ory. In this Figure we see regions of high mobility. as indicated by large RMSF values and regions of much lower mobilit’y. \;Z’e would suggest that, the regic~ns of the structurr with RMSF values less than 1.2 A (which is chosen rather arbtrarily simply to guide our discussions) would repr+ sent conformationally non-labile units while the other we will trrrn conformationally labile. LVit’h
this qualitative classification we find that helic*t+ A. t3, and E to H have regions of non-lability ovtlr these Gmescales whereas helixes (‘ and 1) are Iabilc~. Purt~hermore, helices K. E and (: are thr most M)I~labilr. Helices A and H have regions away from t tic helical core which undrrgo motions but appear as predominantly non-labile hrlical segments. Mot ion in helix F increases steadily from the S-trrminnl rnd into the F(: corner. It is this section of tht, chain t,hat rnovrs into the vacant hemr (*avita?. The tyqion including heticrs (’ and I) shows larger-sc*alc tnotiorrs occurring over’ the timesc.alrs of 300 to 100 ps studied here. Wr notje that’ under t ht relativc4y mild conditions studied here (pH near 7 and tPrnprraturc> around 31dK) these motions represent) c~quilihrium fluctuations. How-ever. we bcllieve these motionh arc’ the indicators of structural regions which an’ “labile” towards unfolding. which would takr place at higher tc~mperaturt~s or a,t lower pH. These motions arc also illustrated by t ht, suprsrposition of the ten picosrc*ond avckrage slruc*t ures shown as in a stereo view of t)heir (‘a traccls in Figurrs 1. We havr chosen structurrs spa~d 10 pic,osec*onds apart, for clarity of t.hr IQgure. From this Figure. regions of t.he molec~ule that arp relativrly notelabile. i.e. those that overlay wf,II in thr super’posiGon, and t,hosr t)hat are labile are pasil). observed. Labile regions such as the, (!-terminal rnd of the R helix t,hrough the N-terminal end of the I< hctlix are c*lrarl>, obsrr\-c>cl in thr Figure with t hta
Figure 4. SuJwrposition of (I” tracrs from average structurrs throughout thp 120 ps molecular dynamiw simulation of is displayed with sampling of one structure apc,myogiot)in. A st,rreo view of structures averaged over 10 ps “windows” r\-rr~’ 40 ps. Thrl large range of longer timescale motions. wJwiall?; for the (1. 1) helical region and FG wrner. is illustrated 1)~ the “scat&r” of structural elements in these regions. In addition. thr stability of thr (: to H wnnrctivr IooJ~. as it J&ks against the A hrllix is noted.
largta range of’ motion present, in thta (‘-1) helical region being outstanding. Similar observations can he made regarding the FG corner with the adjacent. helical termini also displaying motion. Tt is intrresting to note that the non-helical region connecting hrfix G with H shows mohifitSy less than the (‘-I) helical region. This may occur because it packs against the relat,ively non-labile A helix. Toget’her. heficc>s A, C; and H form a subdomain showing little motion in our simulation. This suggests stability of this unit. whicah is in agreement with the hypothesis of Hughson rt ~1. (1990). While it is difficult to make direct comparison with the results from recent n.m.r. experiments (Hughson it N/.. 1990). it is rxt)remefy encouraging that our resuftjs from molecular dynamics calcufat)ions are in good acacbord with the qualitative picture of slruc%ut*e and flexibility in apomyoglobin developed from these experiments. Il’e have identified dXrrenc*rs in t’he motional characterist’ics for regions of thr protein. which we term c-onformationafly labile and non-labile. These regions correspond well with the proton exchange protection factors ot)swved by t1.m.r. (Hughson et al., 1990). a,nd lend support to the suggestion of Hughson rt nl. (1990) that trclliczes A, (: and H might) form a stable folding unit in the I state of apomyoglobin. Furthermore. we fitId tha.t substantial motion owurs in regions adjacent to the home pocket, hefices (‘. I> and F. as well as thr segments directly preceding and following these regions. as a result, of thr absence ot t ht. prosthet ic. group and exposure to solvent. Motions wt’ observed suggest t,hat multiJ)le corlformations ma: contribute to XOE couplings in t flese regions of the sequence and this may compli(~1 e full determination of the solution structure of native apom>-ogfobin hy n.m.r. methods ((I~cc*o & I,ewmte. I 990). fIrsf& our observation t)hat signiticsant motion occurs over timescales in the 200 to 400 ps range. the overall structure atld focal fluctuations we observe for the native state of af)orr~?‘o~fof)il~ are yuitrb similar to t.flose for its hofo
c~ounterpart. This t)oo is in basic agreement with the findings from experimental st,udirs on t.his &em. A final point worth mentioning is the timrscafe required t’o “see” larger-scale mot’ions under condiCons of t,emperahure and pH similar to t,hose being explored in the current n.m.r. experiments. It was necessary to rxbend the molecular dynamic-s simulations beyond the - 100 to %o() [JS timescafe used typicsally for sofvat’ed protein systems of this size. all d we are continuing our analysis of the sirnufations described in this communiration to periods greater than 900 ps. This necessity is a direcat) c‘onsrquence of our desire to observe t,he system near experimental conditions. It is probable t,hat increasing the temperature (Tirado-ISives & .Jorgensen. 1991: IXapua d al., 1990: Fan rt al., 1991: Soman fjt al., 1991; Dagpert &, Levitt. 1992) 01 modifying the nat,ure of the inter-at’ornic* interactions (for example, by making all atoms “see each other” through purely repulsive rlet*trost,atic int’eractions (Sehnften, personal comrnunicaation)) may !,ield a similar pic.t>urr of c~onformat,ionafly labile /‘prsTls non-labile groups. However. since one of our J,rimary objectives is to provide det’a,iled structural models for systems currently under experimental sc*tutiny, we feel that these met’hods are not viable. Instead. we have focussed upon ryuifibrium-based met,hods that permit us to develop nru-native. noncotnJ)act stattbs in aqueous solution. and we will f)ursue the equilibrium studies of i hew novrf stat)es of’ J)art,iafly folded protein in future f)ubficxtions. ‘l’hca author \vc~nltl likta t,o acknowlrdpr thr following agrhncaies fhr Jjartial financial support of’ this work: the Sational TnsCtutes of Health (GM 375.54: 1’4 I RJtMOO9). the Kational Science Foundation (.AL\s(‘X!W?XL)6) and A. J’. Sloan Foundation for fellowship suJ)port. SJw+al t,hanks is owed to I)r *J. Mertz of (:raJ. Resrarc*h. Inc.. for his invaluahlr role in helping with somts of t hr c,alculations rrywrted herein. 1%‘~ are gratrful to the J’ittsburgh SrlJ)rrc~omJ)utiriy (‘rntrl and (?a!, Itcwarc~ti. Inc. for sul)pJying the wmputer rwourws ~~ecwsary to c.arr\- out thrse c~:rl(~ulat~ions. Thr, author woultl also likca tc, thank I)r
3x0
( ‘.
----__-
.\I. 1’clttit.t for elf’ ttolortt?;oplol)in
sharing
his RMSF data solution (Findstw
irl
t’ron~
I,.
I,‘rooks
I I I
_-
hinrulatilltl I!J!ll).
thy
rt ~1..
Iliturn.
.I.
structnrc~
c1vnarnic.s
1'58. Findsrn.
References IZllrn.
M.
I’.
Xinrrrlation
B.
Tildrslr~.
of
Liqrcitls.
Haurn.
(‘. protrin
H. (1973). chain.
I). .J. Oxford
.I ., ( 1989).
I’rinciples
(1989). I’nivctrsity
I)ohson. (‘. (‘hara~terizttt,ion
SMR methods: all)hcc-la~talhumin. Rvrnstrin. Mc~rr.
F. E.
that
181.
A’cirnce.
%I..
the
I’rrss.
A. & folded
C’.. Koetzle. F., Krirr. M.
28.
‘I’. I)..
(iriko.
folding
Hartley. protein
(‘. l)J
guinea
pig
globule state of 7-l
:1.
F.. \Villiamh. Rodgers. .I.
R..
(:. ‘I’. Kinnard.
I
