20 July 1973, Volume 181, Number 4096 6

Principles that Govern theFolding of Protein Chains

Christian B. Anfinsen

The telegram that I received fromthe Swedish Royal Academy of Sciencesspecifically cites ". . . studies on ribo-nuclease, in particular the relationshipbetween the amino acid sequence andthe biologically active conformation.. . . The work that my colleagues andI have carried out on the nature ofthe process that controls the folding ofpolypeptide chains into the uniquethree-dimensional structures of proteinswas, indeed, strongly influenced by ob-servations on the ribonuclease molecule.Many others, including Anson andMirsky (1) in the 1930's and Lumryand Eyring (2) in the 1950's, had ob-served and discussed the reversibilizy ofdenaturation of proteins. However, thetrue elegance of this consequence ofnatural selection was dramatized bythe ribonuclease work, since the refold-ing of this molecule, after full denatura-tion by reductive cleavage of its fourdisulfide bonds (Fig. 1), required thatonly 1 of the 105 possible pairings of

Copyright e 1973 by the Nobel Foundation.The author is chief of the Laboratory of

Chemical Biology, National Institute of Arthritis,Metabolic and Digestive Diseases, National Insti-tutes of Health, Bethesda, Maryland 20014. Thisarticle is the lecture he delivered in Stockholm,Sweden, on 11 December 1972 when he receivedthe Nobel Prize for Chemistry, a prize he sharedwith Stanford Moore and William H. Stein. Itis published here with the permission of theNobel Foundation and will also be included inthe complete volume of Les Prix Nobel en 1972as well as in the series Nobel Lectures (in Eng-lish) published by the Elsevier Publishing Com-pany, Amsterdam and New York. Dr. Moore'sand Dr. Stein's combined lecture appeared as a

single article in the 4 May issue of Science, page458.

20 JULY 1973

eight sulfhydryl groups todisulfide linkages take place.nal observations that led to tsion were made together wileagues Michael Sela and Iin 1956-1957 (3). These wtuality, the beginnings of aof studies that rather vaguelthe eventual total synthesistein. As we all know, Guttefield (4) at the Rockefelhcand Ralph Hirschman and hilat the Merck Research In,have now accomplished thistal task.The studies on the rena

fully denatured ribonucleasmany supporting investigatito establish, finally, the gene;we have occasionally call"thermodynamic hypothesispothesis states that the tsional structure of a nativeits normal physiological milipH, ionic strength, presenccomponents such as metal icthetic groups, temperature,is the one in which theenergy of the whole systenthat is, that the native confdetermined by the totalit3atomic interactions and heamino acid sequence, in a giNment. In terms of naturthrough the "design" of macduring evolution, this ideathe fact that a protein mcmakes stable, structural sei

'CIE:NCES

exists under conditions similar to thosefor which it was selected-the so-calledphysiological state.

After several years of study on theribonuclease molecule it became clearto us, and to many others in the fieldof protein conformation, that proteinsdevoid of restrictive disulfide bonds orother covalent cross-linkages wouldmake more convenient models for thestudy of the thermodynamic and kineticaspects of the nucleation, and subse-quent pathways, of polypeptide chainfolding. Much of what I will reviewdeals with studies on the flexible and

form four convenient staphylococcal nucleaseThe origi- molecule, but I will first summarize

this conclu- some of the older background experi-ith my col- ments on bovine pancreatic ribonucleaseFred White itself.vere, in ac-long seriesly aimed at Support for theof the pro- "Thermodynamic Hypothesis"and Merri-er Institute, An experiment that gave us a partic-is colleagues ular satisfaction in connection withstitute (5), the translation of information in themonumen- linear amino acid sequence into native

conformation involved the rearrange-turation of ment of so-called "scrambled" ribonu-se required clease (8). When the fully reduced pro-:ions (6-8) tein, with eight SH groups, is allowed torality which reoxidize under denaturing conditions[ed (9) the such as exist in a solution of 8 molar." This hy- urea, a mixture of products is obtained,hree-dimen- containing many or all of the possibleprotein in 105 isomeric disulfide bonded forms

ieu (solvent, (schematically shown at the bottom rightce of other of Fig. 2). This mixture is essentiallyons or pros- inactive-having on the order of 1 per-and other) cent the activity of the native enzyme.Gibbs free If the urea is removed and then is lowest; "scrambled" protein is exposed to aformation is small amount of a sulfhydryl group-y of inter- containing reagent such as mercapto--nce by the ethanol, disulfide interchange takesven environ- place, and the mixture eventually is*al selection converted into a homogeneous product,!romolecules indistinguishable from native ribonu-emphasized clease. This process is driven entirelyolecule only by the free energy of conformation thatnse when it is gained in going to the stable, native

223

on

Feb

ruar

y 22

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n F

ebru

ary

22, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Feb

ruar

y 22

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n F

ebru

ary

22, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Feb

ruar

y 22

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n F

ebru

ary

22, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Feb

ruar

y 22

, 201

4w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n F

ebru

ary

22, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

structure. These experiments, inciden-tally, also make unlikely a process ofobligatory, progressive folding duringthe elongation of the polypeptide chain,during biosynthesis, from the NH2- tothe COOH-terminus. The "scrambled"protein appears to be essentially devoidof the various aspects of structural regu-larity that characterize the native mole-cule.A disturbing factor in the kinetics of

the process of renaturation of reducedribonuclease, or of the "unscrambling"experiments described above, was theslowness of these processes, frequentlyhours in duration (7). It had been estab-lished that the time required to syn-thesize the chain of a protein likeribonuclease, containing 124 aminoacid residues, in the tissues of a higherorganism would be approximatelyminutes (10). The discrepancy betweenthe in vitro and in vivo rates led to thediscovery of an enzyme system in theendoplasmic reticulum of cells (particu-larly in those concerned with the secre-tion of extracellular, disulfide-bondedproteins) which catalyzes the disulfideinterchange reaction and which, whenadded to solutions of reduced ribonu-clease or to protein containing random-ized disulfide bonds, catalyzed therapid formation of the correct, nativedisulfide pairing in a period less thanthe requisite 2 minutes (11). The above

discrepancy in rates would not havebeen observed in the case of the fold-ing of structures that were not cross-linked and, as discussed below, suchmotile proteins as staphylococcal nu-clease or myoglobin can undergo virtu-ally complete renaturation in a few sec-onds or less.The disulfide interchange enzyme

subsequently served as a useful tool forthe examination of the thermodynamicstability of disulfide-bonded proteinstructures. This enzyme, having a mo-lecular weight of 42,000 and containingthree half-cystine residues, one of whichmust be in the SH form for activity(12), appears to carry out its rearrang-ing activities on a purely random basis.Thus, a protein whose disulfide bondshave been deliberately broken and re-formed in an incorrect way, need onlybe exposed to the enzyme (with its es-sential half-cystine residue in the pre-reduced, sulfhydryl form) and inter-change of disulfide bonds occurs untilthe native form of the protein substrateis reached. Presumably, disulfide bondsoccupying solvent-exposed, or otherthermodynamically unfavorable posi-tions, are constantly probed and pro-gressively replaced by more favorablehalf-cystine pairings, until the enzymecan no longer contact bonds because ofsteric factors, or because no further netdecrease in conformational free energy

Fig. 1. The amino acid sequence of bovine pancreatic ribonuclease (50).224

can be achieved. Model studies onribonuclease derivatives had shown that,when the intactness of the genetic mes-

sage represented by the linear sequenceof the protein was tampered with bycertain cleavages of the chain, or bydeletions of amino acids at variouspoints, the added disulfide interchangeenzyme, in the course of its "probing,"discovered this situation of thermody-namic instability and caused the randomreshuffling of disulfide bonds with theformation of an inactive cross-linkednetwork of chains and chain fragments[see, for example (13)]. With two natur-ally occurring proteins, insulin andchymotrypsin, the interchange enzymedid, indeed, induce such a randomizingphenomenon (14). Chymotrypsin, con-

taining three disulfide-bonded chains, isknown to be derived from a single-chained precursor, chymotrypsinogen,by excision of two internal bits of se-quence. The elegant studies of Steinerand his colleagues subsequently showedthat insulin was also derived from a

single-chained precursor, proinsulin(Fig. 3), which is converted tothe two-chained form, in whichwe normally find the active hor-mone, by removal of a segment fromthe middle of the precursor strand afterformation of the three disulfide bonds(15). In contrast, the multichained im-munoglobulins are not scrambled andinactivated by the enzyme, reflectingthe fact that they are normal productsof the disulfide bonding of four pre-formed polypeptide chains.

Factors Contributing to the CorrectFolding of Polypeptide Chains

The results with the disulfide inter-change enzyme discussed above sug-gested that the correct and uniquetranslation of the genetic message for aparticular protein backbone is no longerpossible when the linear information hasbeen tampered with by deletion ofamino acid residues. As with most rules,however, this one is susceptible tomany exceptions. First, a number ofproteins have been shown to undergoreversible denaturation, including disul-fide bond rupture and re-formation,after being shortened at either theNHW- or COOH-terminus (16). Othersmay be cleaved into two (17-19), oreven three, fragments which, althoughdevoid of detectable structure alone insolution, recombine through noncova-

SCIENCE, VOL. 181

lent forces to yield biologically activestructures with physical properties verysimilar to those of the parent proteinmolecules. Richards and his colleagues(17) discovered the first of these recom-bining systems, ribonuclease S, whichconsists of a 20-residue fragment fromthe NH2-terminal end held by a largenumber of noncovalent interactions tothe rest of the molecule which, in turn,consists of 104 residues and all fourof the disulfide bridges. The work byWyckoff, Richards, and their associateson the three-dimensional structure ofthis two-fragment complex (20) andstudies by Hofmann (21) and Scoffone(22) and their colleagues on semisyn-thetic analogs of this enzyme deriva-tive are well known. Studies in our ownlaboratory (23) showed that the 20-residue "ribonuclease S peptide" frag-ment could be reduced by five residuesat its COOH-terminus without loss ofenzymic activity in the complex, or ofits intrinsic stability in solution.

Other examples of retention of nativestructural "memory" have been foundwith complexing fragments of thestaphylococcal nuclease molecule (18,24). This calcium-dependent, RNA-and DNA-cleaving enzyme (Fig. 4)consists of 149 amino acids and isdevoid of disulfide bridges and sulfhy-dryl groups (25). Although it exhibitsconsiderable flexibility in solution, asevidenced by the ready exchange oflabile hydrogen atoms in the interiorof the molecule with solvent hydrogenatoms (26), only a very small fractionof the total population deviates from theintact, native format at any moment.Spectral and hydrodynamic measure-ments indicate marked stability up totemperatures of approximately 55°C.The protein is greatly stabilized, bothagainst hydrogen exchange (26) andagainst digestion by proteolytic enzymes(27) when calcium ions and the inhibi-tory ligand, 3',5'-thymidine diphosphate(pdTp), are added. Trypsin, for ex-ample, only cleaves at very restrictedpositions-the loose NH2-terminal por-tion of the chain and a loop of residuesthat protrudes out from the molecule asvisualized by x-ray crystallography.Cleavage occurs between lysine residues5 and 6 and, in the sequence -Pro-Lys-Lys-Gly- (residues 47 through 50) (28),between residues 48 and 49 or 49 and50 (18). The resulting fragments (resi-dues 6 to 48) and (49 to 149) or (50 to149), are devoid of detectable structurein solution (29); however, as in the case

20 JULY 1973

in vitro

Fig. 2. Schematic representation of thereductive denaturation, in 8M urea solu-tion containing 2-mercaptoethanol, of adisulfide-cross-linked protein. The con-version of the extended, denatured formto a randomly cross-linked, "scrambled"set of isomers is depicted at the lowerright.

Fig. 3. The structure of porcine proinsulin (51).

LY -COOH

Fig. 4. Covalent structure of the major extracellular nuclease of Staphylococcusaureus (25).

225

100 200Volume of eluate(ml)

0W

3

_

Fig. 5. Isolation of semisynthetic nu-clease T on a phosphocellulose columnafter "functional" purification by trypsindigestion in the presence of calcium ionsand thymidine-3'5'-diphosphate (35).

of ribonuclease S, when these structure-less fragments are mixed in stoichio-metric amounts, regeneration of activity(about 10 percent) and of native struc-tural characteristics occurs (the com-plex is called nuclease T). Nuclease Thas now been shown (30) to be closelyisomorphous with native nuclease (31).Thus the cleavages and deletions do notdestroy the geometric "sense" of thechain. Recently it was shown that resi-due 149 may be removed by carboxy-peptidase treatment of nuclease, and

l, I I I

U,

60-

050-

0 40 -

30

to 30te e thSemisyntheticnuclease Tlel

20

10 Nuctease T (494)

300 320 340 360 380 400Wavelength of emission(nm)

Fig. 6. Use of fluorescence measurementsto determine the relative hydrophobicity(presumably reflecting "nativeness" in thecase of nuclease) of the molecular en-vironment of the single tryptophan residuein this protein (33, 35).

that residues 45 through 49 are dispens-able, the latter conclusion the result ofsolid phase-synthetic studies (32) onanalogs of the fragment (6 to 47).

Earlier studies by David Ontjes (33)had established that the rapid and con-venient solid-phase method developedby Merrifield (34) for peptide synthesiscould be applied to the synthesis ofanalogs of the fragment (6 to 47) of

nuclease T. The products, although con-taminated by sizable amounts of "mis-take sequences" that lack amino acidresidues because of slight incomplete-ness of reaction during coupling, couldbe purified by ordinary chromatographicmethods to a stage that permitted oneto make definite conclusions about therelative importance of various compo-nents in the chain. Taking advantage ofthe limited proteolysis that occurs whennuclease is treated with trypsin in thepresence of the stabilizing ligands, cal-cium and pdTp, Chaiken (35) was ableto digest away those aberrant syntheticmolecules of (6 to 47) that did notform a stable complex with the large,native fragment (49 to 149). After di-gestion of the complex, chromatographyon columns of phosphocellulose (Fig.5) yielded samples of semisyntheticnuclease T that were essentially indis-tinguishable from native nuclease T.For example, the large enhancement offluorescence of the single tryptophanresidue in nuclease [located at position140 in the fragment (50 to 149)] uponaddition of the native fragment (6 to49) was also shown when, instead,synthetic (6 to 47) peptide isolatedfrom semisynthetic nuclease T that hadbeen purified as described above wasadded (Fig. 6).The dispensalbility, or replaceability,

of a number of residues to the stabilityof the nuclease T complex was estab-lished by examining the fluorescence,activity, and stability to enzymaticdigestion of a large number of semisyn-thetic analogs (36). As is illustrated inFig. 7, interaction with the calciumatom required for nuclease activity nor-

Fig. 7 (left). Amino acid residues in the sequence of nucleasethat are of particular importance in the catalytic activity andbinding of substrate and calcium ions (36). Fig. 8 (right).A schematic view of the three-dimensional structure of staph-ylococcal nuclease (31, 47) .

SCIENCE, VOL. 181

goN

226

mally requires the participation of fourdicarboxylic amino acids. Although theactivities of complexes conaining syn-thetic (6 to 47) fragments in which oneof -these had been replaced with anasparagine or glutamine residue wereabolished (with one partial exception-asparagine at position 40), three-dimensional structure and complexstability were retained for the mostpart. Similarly, replacement of arginineresidue 35 with lysine yielded an inac-tive complex, but nevertheless one withstrong three-dimensional similarity tonative nuclease T.A second kind of complementing

system of nuclease fragments (24) con-sists of tryptic fragment (1 to 126) anda partially overlapping section of thesequence (99 to 149) prepared bycyanogen bromide treatment of thenative molecule (shown schematicallyin Fig. 8). These two peptides form acomplex with about 15 percent of theactivity of nuclease itself, which is suf-ficiently stable in the presence of pdTpand calcium ions to exhibit remarkableresistance to digestion by trypsin. Thus,many of the overlapping residues inthe complex consisting of (1 to 126):(99 to 149), may be "trimmed" awaywith the production of a derivative, (1to 126): (1l1 to 149). Further degrada-tion of each of the two components, theformer with carboxypeptidases A and Band the latter with leucine aminopepti-dase, permits the preparation of (1 to124): (114 to 149), which is as activeand as structurally similar to nativenuclease (as evidenced by estimates ofhydrodynamic, spectral, and helicalproperties) as the parent, undegradedcomplex. A number of synthetic analogsof the (114 to 149) sequence have beenprepared (37), which also exhibit ac-tivity and "native" physical propertieswhen added to (1 to 126). I will discussbelow the manner in which these com-plexing fragments have been useful indevising experiments to study the pro-cesses of nucleation and folding ofpolypeptide chains.

Mutability of Information for

Chain Folding

Biological function appears to bemore a correlate of macromoleculargeometry than of chemical detail. Theclassic chemical and crystallographicwork on the large number of abnormalhuman hemoglobins, on the species var-iants of cytochrome c, and on other

20 JULY 1973

100 F

I

80

40

m.

tl, Reducd vicosty

* Molar ellIpticity, 220m

a

.a

& a u L-

I I I I _

IF

20 _-

0

2 3 4 a 6

PH7

Fig. 9. Changes in reduced viscosity andmolar ellipticity at 220 nm during theacid-induced transition from native todenatured nuclease. 0 and U, Reducedviscosity; A and A, molar ellipticity at220 nm; and measurements madeduring the addition of acid; * and A,measurements made during the addition ofbase (44).

proteins from a very large variety ofsources and the isolation of numerous

bacterial proteins after mutation of thecorresponding genes have made itquite clear that considerable modifica-tion of protein sequence may be madewithout loss of function. In those caseswhere crystallographic studies of three-dimensional structure have been made,the results indicate that the geometricproblem of "designing," through nat-ural selection, molecules that can sub-serve a particular functional need can

be solved in many ways. Only thegeometry of the protein and its activesite need be conserved, except, ofcourse, for such residues as actuallyparticipate in a unique way in a catalyt-ic or regulatory mechanism (38). Studiesof model systems have led to similarconclusions. In our own work on ri;bo-nuclease, for example, it was shownthat fairly long chains of poly-DL-alanine could be attached to eight of theeleven amino groups of the enzymewithout loss of enzyme activity (39).Furthermore, the polyalanylated enzymecould be converted to an extendedchain by reduction of the four disulfidebridges in 8M urea, and this fully dena-tured material could then be reoxidizedto yield the active, correctly foldedstarting substance. Thus, the chemistryof the protein could be greatly modi-fied, and its capacity to refold afterdenaturation seemed to be dependentonly on internal residues and not thoseon the outside, exposed to solvent. This

is, of course, precisely the conclusionreached by Perutz and his colleagues(40), and by others (41) who have re-viewed and correlated the data on vari-ous protein systems. Mutation andnatural selection are permitted a highdegree of freedom during the evolutionof species, or during accidental muta-tion, but a limited number of residues,destined to become involved in the in-ternal, hydrophobic core of proteins,must be carefully conserved (or at mostreplaced with other residues with aclose similarity in bulk and hydropho-bicity).

Cooperativity Required forFolding and Stability of Protein

The examples of noncovalent inter-action of complementing fragments ofproteins quoted above give strong sup-port to the idea of the essentiality ofcooperative interactions in the stabilityof protein structure. As is the case inany language, an incomplete sentencefrequently conveys only gibberish. Thereappears to exist a very fine balance be-tween stable, native protein structureand random, biologically meaninglesspolypeptide chains.A very good example of the inade-

quacy of an incomplete sequence comesfrom our observations on the nucleasefragment (1 to 126). This fragment con-tains all of the residues that make upthe active center of nuclease. Neverthe-less, this fragment, representing about85 percent of the total sequence of nu-clease, exhibits only about 0.12 per-cent of the activity of the native en-zyme (42). The further addition of 23residues during biosynthesis, or the ad-dition in vitro, of residues 99 to 149 asa complementing fragment (24), re-stores the stability required for activityto this unfinished gene translation.The transition from incomplete, in-

active enzyme, with random structure,to competent enzyme, with unique andstable structure, is clearly a delicatelybalanced one. The sharpness of thistransition may be emphasized by ex-periments of the sort illustrated in Fig.9. Nuclease undergoes a dramaticchange from native globular structureto random disoriented polypeptide overa very narrow range of pH, centeredat pH 3.9. The -transition has the ap-pearance of a "two-stage" process-either all native or all denatured-and,indeed, two-state mathematical treat-ment has classically been employed to

227

120. *

JL..

L-

"'A

Table 1. Studies of the equilibrium between the peptide fragment (99 to 149) in its randomform [fragment (99 to 149),r and in the form thi, fragment assumes in the native structure ofnuclease [fragment (99 to 149)j]. Abbreviations: P, fragment (99 to 149); Ab, antibody; conf,conformation; assoc, association; T, total; t, time.

[AbP.]Ka>>oc [Ab] [PT]

[AbItotal sites [PT] t112 [Ab]fitrest [Ab]bound sites K PT aS P,,(,uM) (,uM) (sec) (A4M) (,uM)0.076 0 18 0.076 0.076 0.65 20 .068 0.0080 2.20 X 10-4 0.022.076 2.0 24 .057 .019 2.02 x 10-4 .020.076 2.6 27 .051 .025 2.29 X 10-4 .023.076 7.8 35 .039 .037 1.47 X 10-4 .015.076 6.5 33 .042 .034 1.51 X 10-4 .015

K,=.f= (2.0 ± 0.4) X 10-4

describe such data. In actuality, it hasbeen possible to show, by nuclear mag-netic resonance and spectrophotometric-experiments (43) that one of the fourhistidines and one tyrosine residue ofthe seven in nuclease become disori-ented before the general and suddendisintegration of organized structure.However, such evidences of a stepwisedenaturation and renaturation processare certainly not typical of the bulkof the cooperatively stabilized mole-cule.The experiments in Fig. 9, involv-

ing measurements of intrinsic viscosityand helix-dependent circular dichro-ism, are typical of those obtained withmost proteins. In the case of nuclease,not only is the transition from nativeto denatured molecule during transferfrom solution at pH 3.2 to 6.7 veryabrupt, but the process of renaturationoccurs over a very short time period. Ishall not discuss these stop-flow kineticexperiments (44) in detail. In brief, theprocess can be shown to take place inat least two phases-an initial rapidfolding with a half-time of about 50milliseconds and a second, somewhatslower transformation with a half-timeof about 200 milliseconds. The first

NATIVE%N -1FORMAT

I~1

NATIVEHOO OC FORMATC_

141

NATIVE NATIVEFORMAT FORMAT COOH

I

IL

phase is essentially temperature-inde-pendent (and therefore possibly en-tropically driven) and the second tem-perature-dependent.

Nucleation of Folding

A chain of 149 amino acid residuesvith two rotatable bonds per residue,each bond probably having two orthree permissible or favored orienta-tions, would be able to assume on theorder of 4149 to 9149 different confor-mations in solution. The extreme ra-pidity of the refolding makes it essen-tial that the process take place along alimited number of "pathways," evenwhen the statistics are severely restrict-ed by the kinds of stereochemicalground rules that are implicit in a so-called Ramachandran plot. It becomesnecessary to postulate the existence ofa limited number of allowable initiatingevents in the folding process. Suchevents, generally referred to as nucle-ations, are most likely to occur in partsof the polypeptide chain that can par-ticipate in conformational equilibriabetween random and cooperativelystabilized arrangements. The likelihood

Fig. 10 (left). How protein chainsmight fold (see the text for a dis-cussion of this fairly reasonable,but subjective proposal). Fig.11 (right). Inhibition of nucleaseactivity by antibody to (99 to149),, and lack of inhibition byantibody to (99 to 149), madeagainst the peptide (99 to 149),presumably in a random confor-mation (45).

of a requirement for cooperative sta-bilization is high because, in aqueoussolution, ionic or hydrogen-bonded in-teractions would not be expected tocompete effectively with interactionswith solvent molecules and anythingless than a sizable nucleus of interact-ing amino acid side chains would prob-ably have a very short lifetime. Further-more, it is important to stress that theamino acid sequences of polypeptidechains designed to be the fabric of pro-tein molecules only make functionalsense when they are in the three-di-mensional arrangement that charac-terizes them in the native protein struc-ture. It seems reasonable to suggest thatportions of a protein chain that canserve as nucleation sites for foldingwill be those that can "flicker" in andout of the conformation that they oc-cupy in the final protein, and that theywill form a relatively rigid structure,stabilized by a set of cooperative inter-actions. These nucleation centers, inwhat we have termed their "native for-mat" (Fig. 10), might be expected toinvolve such potentially self-dependentsubstructures as helices, pleated sheets,or ibeta-bends.

Unfortunately, the methods that de-pend on hydrodynamic or spectral mea-surements are not able to detect thepresence of these infrequent and tran-sient nucleations. To detect the postu-lated "flickering equilibria" and to de-termine their probable lifetimes in so-lution requires indirect methods thatwill record the brief appearance of in-dividual "native format" molecules inthe population under study. One suchmethod, recently used in our labora-tory in a study of the folding of staphy-lococcal nuclease and its fragments, em-ploys specific antibodies against re-

0.40.

00cm

1.20

0 1 2 3 4Time (minutes)

SCIENCE, VOL. 181

1I.

l

No Antibody

Anti(99.149)r.......006........

Anti(99.149)n ..

Nuclease0.04Jug

228

LOUl _I

,.uv_

stricted portions of the amino acid se-quence (45).

Figure 8 depicts the three-dimension-al pattern assumed by staphylococcalnuclease in solution. Major features in-volving organized structure are thethree-stranded antiparallel pleated sheetapproximately located between resi-dues 12 and 35, and the three alpha-helical regions between residues 54 to67, 99 to 106, and 121 to 134. An-tibodies against specific regions of thenuclease molecule were prepared byimmunization of goats with either poly-peptide fragments of the enzyme or byinjection of the intact, native proteinwith subsequent fractionation of theresulting antibody population on affin-ity chromatography columns consistingof agarose bearing the covalently at-tached peptide fragment of interest (45,46). In the former manner there wasprepared, for example, an antibody di-rected against the polypeptide, residues99 to 149, known to exist in solutionas a random chain without the exten-sive helicity that characterizes this por-tion of the nuclease chain when pres-ent as part of the intact enzyme. Suchan antibody preparation is referred toas anti-(99 to 149),r the subscript in-dicating the disordered state of theantigen.When, on the other hand, a fraction

of antiserum to native nuclease, iso-lated on an agarose-nuclease column,was further fractionated on agarose-fragment (99 to 149), a fraction was ob-tained which was specific for the se-quence (99 to 149), but presumably onlywhen this bit of sequence occupied the"native format." This latter conclusionis based on the observation that thelatter fraction, termed anti-(99 to 149)n(the subscript n referring to the nativeformat) exhibited a strong inhibitoryeffect on the enzymic activity of nu-clease (47) whereas anti-(99 to 149)rwas devoid of such an effect (see Fig.11). This conclusion was further sup-ported by the observation that the con-formation-stabilizing ligands, pdTp andcalcium ions, showed a marked inhibi-tory effect on the precipitability of nu-clease by anti-(99 to 149)r but had lit-tle effect, if any, on such precipitabilityby anti-(1 to 149),, (45). This findingreinforced the idea that many of theantigenic determinants recognized bythe antibodies to fragments are presentonly in the "unfolded" or "nonnative"conformation of nuclease. A furthersubfractionation, yielding anti-(99 to126),,, was carried out by passage of

20 JULY 1973

Fig. 12. Semilogarithmicplot of activity againsttime for assays of 0.05#Ag of nuclease in thepresence of: 0-0, noantibody; *-@, 6 ugof antibody to (99 to126),; 0-O, 6 /Ag ofantibody to (99 to 126)nplus 12 jug of (99 to149); and A-A, 6 ,tgof antibody to (99 to126)n plus 48 ,Ag of (99to 149). The dotted linerepresents one-half of theinitial activity.

c

E

tocm

N

:

-o

_-

anti-(99 to 149). through a column ofSepharose-immobilized fragment (127to 149). This antibody fraction, whichforms an inactive and soluble complexwith nuclease, was used in the experi-ments described below. The reactionbetween anti-(99 to 149) and nucleasecould be shown by measurements ofchanges in the kinetics of inhibitionof enzyme activity (Fig. 12) to be ex-tremely rapid, with k,,,, = 4.1 X 1 0"M-sec-1; k0ff, on the other hand, is small.The system may be described by two

simultaneous equilibria, the first con-cerned with the "flickering" of frag-ment (99 to 149), which we shall termP, from random to native "format,"and the second with the association ofanti-(99 to 126),1, which we shall term,simply, Ab with fragment P in its na-tive format: that is, P,,.

[Pi]Ab + PI,= AbP,. KA1S. =-(AbPoll

[Ab] [Psi]Keollt- [AbP,ll

K....oe [Ab] [P.,]The amount of unbound antibodyin the second equilibrium maybe estimated from measurementsof the kinetics of inactivation ofthe digestion of denatured DNA sub-strate by a standard amount of nucleaseadded to the previously incubated mix-ture of fragment (99 to 149) and anti-(9 to 126),,. If we make the assumptionthat the affinity of anti-{99 to 126). forfragment (99 to 149) in its folded (P)form is the same as that determined forthis antigenic determinant in nativenuclease, the value for the term K,,,,f

may be calculated from measurableparameters. A series of typical valuesshown in Table 1, suggests that approxi-mately 0.02 percent of fragment (99 to149) exists in the native format at anymoment. Such a value, although low, isprobably very large relative to the likeli-hood of a peptide fragment of a pro-tein being found in its native formaton the basis of chance alone.

Empirical considerations of the Iargeamount of data now available on cor-relations between sequence and three-dimensional structure (48), together withan increasing sophistication in thetheoretical treatment of the energeticsof polypeptide chain folding (49) arebeginning to make more realistic theidea of the a priori prediction of pro-tein conformation. It is certain thatmajor advances in the understanding ofcellular organization, and of the causesand control of abnormalities in such or-ganization, will occur when we canpredict, in advance, the three-dimen-sional phenotypic consequences of agenetic message.

References and Notes

1. M. L. Anson, Advan. Protein Chem. 2, 361(1945).

2. R. Lumry and H. Eyring, J. Phys. Chem. S8,110 (1954); C. H. W. Hirs, S. Moore, W. H.Stein, J. Biol. Chem. 235, 633 (1960); J. T.Potts, A. Berger, J. Cooke, C. B. Anfinsen,Ibid. 237, 1851 (1962); D. G. Smyth, W. H.Stein, S. Moore, Ibid. 238, 227 (1963).

3. M. Sela, F. H. White, C. B. Anfinsen, Science125, 691 (1957).

4. B. Gutte and R. B. Merrifield, J. Biol. Chem.246, 1922 (1971).

5. R. Hirschmann, R. F. Nutt, D. F. Veber, R.A. Vitali, S. L. Varga, T. A. Jacob, F. W.Holly, R. G. Denkewalter, J. Amer. Chem.Soc. 91, 507 (1969).

6. F. H. White, Jr., and C. B. Anfinsen, Ann.N.Y. Acad. Sci. 81, 515 (1959); F. H. White,Jr., J. Biol. Chem. 236, 1353 (1961).

7. C. B. Anfinsen, E. Haber, M. Sela, F. H.

229

White, Jr., Proc. Nat. Acad. Sci. U.S.A. 47,1309 (1961).

8. E. Haber and C. B. Anfinsen, J. Bfol. Chem.237, 1839 (1962).

9. C. J. Epstein, R. F. Goldberger, C. B. An-finsen, Cold Spring Harbor Symp. Quant.Biol. 28, 439 (1963).

10. H. M. Dintzis, Proc. Nat. Acad. Sci. U.S.A.47, 247 (1961); R. E. Canfield and C. B. An-finsen, Biochemistry 2, 1073 (1963).

11. R. F. Goldberger, C. J. Epstein, C. B. Anfin-sen, J. Biol. Chem. 238, 628 (1963); P. Vene-tianer and F. B. Straub, Biochim. Biophys.Acta 67, 166 (1963).

12. F. DeLorenzo, R. F. Goldberger, E. Steers,D. Givol, C. B. Anfinsen, J. Biol. Chem.241, 1562 (1966); S. Fuchs, F. DeLorenzo, C.B. Anfinsen, fbid. 242, 398 (1967).

13. I. Kato and C. B. Anfinsen, ibid., 244, 5849(1969).

14. G. Givol, F. DeLorenzo, R. F. Goldberger,C. B. Anfinsen, Proc. Nat. Acad. Sci. U.S.A.53, 766 (1965).

15. D. F. Steiner, Trans. N.Y. Acad. Sci. Ser.2 30, 60 (1967).

16. C. B. Anfinsen, Dev. Biol. Suippl. 2, 1 (1968).17. F. M. Richards, Proc. Nat. Acad. Sci. U.S.A.

44, 162 (1958).18. H. Taniuchi, C. B. Anfinsen, A. Sodja, ibid.

58, 1235 (1967).19. I. Kato and N. Tominaga, Fed. Eur. Biochem.

Soc. Lett. 10, 313 (1970).20. H. W. Wyckoff, D. Tsernoglou, A. W. Han-

son, J. R. Knox, B. Lee, F. M. Richards, J.Biol. Chem. 245, 305 (1970).

21. K. Hofmann, F. M. Finn, M. Linetti, J.Montibeller, G. Zanetti, J. Amer. Chem. Soc.88, 3633 (1966).

22. E. Scoffone, R. Rocchi, F. Marchiori, L.Moroder, A. Marzotto, A. M. Tamburro,ibid. 89, 5450 (1967).

23. J. T. Potts, Jr., D. M. Young, C. B. Anfinsen,J. Biol. Chem. 238, 2593 (1963).

24. H. Taniuchi and C. B. Anfinsen, ibid. 246,2291 (1971).

25. J. L. Cone, C. L. Cusumano, H. Taniuchi,C. B. Anfinsen, ibid., p. 3103; J. L. Bohnertand H. Taniuchi, ibid. 247, 4557 (1972).

26. A. N. Schechter, L. Moravek, C. B. Anfinsen,ibid. 244, 4981 (1969).

27. H. Taniuchi, L. Moravek, C. B. Anfinsen,ibid., p. 4600.

28. The abbreviations for the amino acid residuesare: Ala, alanine; Arg, arginine; Asp, as-partic acid; Cys, half-cystine; Glu, glutamicacid; Gly, glycine; His, histidine; Ileu, iso-leucine; Leu, leucine; Lsy, lysine; Met, meth-ionine; Phe, phenylalanine; Pro, proline; Ser,serine; Thr, threonine; Tyr, tyrosine; Val,valine; Glu-NH2, glutamine; Asp-NH2, as-paragine.

29. H. Taniuchi and C. B. Anfinsen, J. Biol.Chem. 244, 3864 (1969).

30. H. Taniuchi, D. Davies, C. B. Anfinsen,ibid. 247, 3362 (1972).

31. A. Arnone, C. J. Bier, F. A. Cotton, E. E.Hazen, Jr., D. C. Richardson, J. S. Richard-son, A. Yonath, ibid. 246, 2302 (1971).

32. G. R. Sanchez, I. M. Chaiken, C. B. An-finsen, ibid., in press.

33. D. Ontjes and C. B. Anfinsen, ibid. 244, 6316(1969).

34. R. B. Merrifield, Science 150, 178 (1965).35. I. M. Chaiken, J. Biol. Chem. 246, 2948

(1971).36. ~ and C. B. Anfinsen, ibid., p. 2285.37. I. Parikh, L. Corley, C. B. Anfinsen, ibid.,

p. 7392.38. W. M. Fitch and E. Margoliash, El'ol. Biol.

4, 67 (1970).39. J. P. Cooke, C. B. Anfinsen, M. Sela, J. Biol.

Chem. 238, 2034 (196V1.40. M. F. Perutz, J. C. Kendrew, H. C. Watson,

J. Mol. Biol. 13, 669 tlvo3).41. C. J. Epstein, Nature 210, 25 (1966).42. D. Sachs, H. Taniuchi, A. N. Schechter, A.

Eastlake, unpublished work.43. H. Epstein, A. N. Schechter, J. Cohen, Proc.

Nat. Acad. Sci. U.S.A. 68, 2042 (1971).44. H. F. Epstein, A. N. Schechter, R. F. Chen,

C. B. Anfinsen, J. Mol. Biol. 60, 499 (1971).45. D. H. Sachs, A. N Schechter, A. Eastlake,

C. B. Anfinsen, Proc. Nat. Acad. Sci. U.S.A.69, 3790 (1972).

46. , J. Immunol. 109, 1300 (1972).47. , Biochemistry 11, 4268 (1972).48. C. B. Anfinsen and H. Scheraga, Advan.

Prot. Chem., in preparation.49. H. A. Scheraga, Chem. Rev. 71, 195 (1971).50. C. H. W. Hirs, S. Moore, W. H. Stein, J.

Biol. Chem. 235, 633 (1960); J. T. Potts, A.Berger, J. Cooke, C. B. Anfinsen, ibid., 237,1851 (1962); D. G. Smyth, W. H. Stein, S.Moore, ibid., 238, 227 (1963).

51. R. E. Chance, R. M. Ellis, W. W. Bromer,Science 161, 165 (1968).

Methylmercury Poisoning in Iraq

An interuniversity report.

F. Bakir, S. F. Damluji, L. Amin-Zaki, M. Murtadha,A. Khalidi, N. Y. Al-Rawi, S. Tikriti, and H. I. Dhahir

T. W. Clarkson, J. C.

Methylmercury and its short-chainhomolog elicit characteristic toxic ef-fects in man that differ from the toxiceffects of other mercury compounds(la, 2). The primary signs and symp-toms of methylmercury poisoning re-sult from damage to the nervous systemand are characterized by loss of sensa-tion at the extremities of the hands

The first eight authors are members of theUniversity of Baghdad, Baghdad, Iraq. Dr. Tikritiis also a member of the Directorate of PreventiveMedicine, Ministry of Health, Iraq. Drs. Clarkson,Smith, and Doherty are in the departments ofradiation biology and biophysics, pharmacologyand toxicology, and pediatrics, respectively, ofthe University of Rochester School of Medicineand Dentistry, Rochester, New York 14642. See(1) for an acknowledgment of the substantial con-tributions that many others have made to thisstudy.

230

(University of Baghdad)Smith, and R. A. Doherty(University of Rochester)

and feet and in areas around the mouth(paresthesia), loss of coordination ingait (ataxia), slurred speech (dysarth-ia), diminution of vision (concentricconstriction of the visual field), andloss of hearing. Severe poisoning can

cause blindness, coma, and death.There is a latent period of weeks or

months between exposure to methyl-mercury and the development ofpoisoning symptoms. Ataxia may sub-sequently decrease but general recoveryis poor. Prenatal exposure to methyl-mercury has resulted in mental retarda-tion with cerebral palsy. Prior to thepresent outbreak in Iraq, between 200and 300 cases of methylmercurypoisoning had been reported in Iraq

and in other parts of the world andmore than 1000 cases had been as-cribed to exposure to the ethylmercuryhomolog (la, 2). The earliest caseswere due to occupational exposure fol-lowing the introduction of methylmer-cury compounds as antifungal seeddressing agents. In the 1950's, reportsof poisonings from nonoccupationalsources appeared in the literature withincreasing frequency. These included afew cases arising from the treatmentof fungal skin infections as well asaccidental and suicidal ingestion. Sev-eral large incidents of poisoning haveoccurred in Iraq, Pakistan, and Guate-mala due to the ingestion of flour andwheat seed treated with methyl- andethylmercury compounds. The fungi-cide ethylmercury-p-toluene sulfonani-lide was claimed to be responsible fortwo outbreaks in Iraq in 1956 and1960.

In the 1960 outbreak, an estimated1000 patients were affected by methyl-mercury poisoning and 370 were ad-mitted to hospital. In Guatemala, casesthat were originally thought to be viralencephalitis were reported during thewheat growing seasons of 1963, 1964,and 1965. Forty-five people were af-fected and 20 died. Methylmercury di-cyandiamide, used to treat the seedwheat before distribution to farmers,was later established as the causativeagent. A similar outbreak occurred inPakistan in 1969.

Despite the large number of people

SCIENCE, VOL. 181