;::-4

0)(11-Yí

C LOO~

The plasma protein Librinogen is con verted into Librin to form a

blood cloro In time the clot breaks down. Both processes can now

be understood in terms oi the detailed structure oi the protein

by Russell F. Doolittle

c

called on. It gives rise to the enzymeplasmin, which destroys fue fibrin net-work and restores the fluidity of fueplasma. Now the question becomes:What kind of molecule is it that can po-lymerize on being attacked by one pro-tease (thrombin) and then "depolymer-ize" under attack by another protease(plasmin)? Put another way, how is fuefunctioning of fibrinogen and fibrin ex-plained by their structure? The answer isthe subject of this article.

The str,ucture of fibrinogen has beenhighly controversial for fue past 10 or15 years. Although some skirmishingcontinues among a few protein chem-ists, 1 think there is now overwhelmingevidence, from a variety of investigativetechniques, to support the structure 1shall presento

O ne so urce of controversy has beenapparently conflicting evidence

from electron micrography. In 1959 Ce-cil E. Hall and Henry S. Slayter, whowere working at fue Massachusetts In-stitute of Technology, published exquis-ite micrographs that showed fibrino-gen as a trinodular, or three-globule,structure [see "The Clotting of Fibrin-ogen," by Koloman Laki; SCIENTIFICAMERICAN, March, 1962]. Unfortunate-ly structural chemists were a long waybehind fue electron microscopists at fuetime, and although fue trinodular formwas for a while generally accepted, therewas little in fue way of chemical un-derpinning for it. As it happened, fuemethodology of electron microscopychanged rapidly in the early 1960's. Halland Slayter had made their micrographswith the shadowing technique, and thennegative staining carne to be the pre-ferred method. Fibrinogen does not takewell to most negative stains, and a seriesof reports over the next decade held thatfue trinodular structure observed byHall and Slayter was merely an arti-Cacto Instead, these later workers main-tained, fibrinogen was more likea bloat-ed beach ball, which opened up duringpolymerization to form fibrin. Recent-

P rick us and we bleed, but the bleed-ing stops; the blood clots. Thesticky cell fragments called plate-

lets clump at fue site of fue puncture,partially sealing fue leak. The plateletshave another important function: theirsurface serves as an assembly afea forthe mobilization of an elaborate force ofenzymes, fue ultimate product of whichis fue protein-cleaving enzyme throm-bin. The target for that protease is fibrin-ogen, a large molecule that ordinarilycirculates, dissolved, in fue blood plas-ma. Under attack by thrombin fue fi-brinogen undergoes a remarkable trans-formation. The altered molecules linkup, spontaneously aligning themselvesinto the long threadlike polymer calledfibrin, fue prirnary ingredient of a bloodclot. The strands of fibrin form a net-work. The network (which forms only inthe direct vicinity of the clumped plate-lets, because that is whe~e thrombin isgenerated) pervades fue blood plasmalocally and changes it from a solution toa gel. The blood clots.

What kind of molecule is fibrinogenthat it can be attacked by a pro tease(thrombin) and thereupon spontaneous-ly'polymerize? That question should notbe dealt with until another aspect of fueclotting phenomenon is considered. Afibrin clot is designed not to last forever.It is a transitory device, quickly formedand soon dismantled in favor of a morepermanent repair system. There areeven more pressing reasons for dissolv-ing a blood clot. A small piece of a clot(an embolus) dislodged into fue circula-tory system can occlude a small bloodvessel; an embolism in the lung can befatal. A thrombus-a clot formed at thewrong time and place, not in response toinjury but as a consequence of some dis-ease process such as atherosclerosis-can lead to a heart attack or a stroke.

To minirnize fue danger of embolismand to reduce the likelihood of throm-bosis, as well as to clear away a thera-peutic clot in the wound-healing proc-ess, a proteolytic system complementa-ry to the one that generates thrombin is

c

ly, however, shadowing techniques havemade a comeback, enabling a number oflaboratories once again to show a tri-nodular structure for fibrinogen. More-over, several workers have even beenable to visualize a trinodular structureby means of negative staining.

Many other experiments support a tri-nodular (or at least a polydomainal)structure for fibrinogen, particularlywhen they are all taken together. Theseinclude an early X-ray-diffraction study,a variety of hydrodynamic measure-ments (such as analytical ultracentrifu-gation) and other physicochemical de-terminations, and also studies in whichfue fibrinogen molecule is fragmentedwith enzymes. In recent years fue entirecovalent structure of fue human fibrino-gen molecule (the full sequence of theamino acid units of its protein chainsand fue ways fue chains are intercon-nected) has been established. It con-firms in detail a particular polydomain-al structure.

In fue 1930's W. T. Astbury of fueUniversity of Leeds undertook an inten-sive study of fue structure of fibrousproteins by fue method of X-ray diffrac-tion. In a fibrous material the moleculesare presumably aligned lengthwise, sothat fue repeat dist¡mce from one aminoacid to another, if it is regular, is re-vealed by the scattered X rays when thebeam is perpendicular to fue fiber axis.In fue early 1940's Kenneth Bailey col-laborated with Astbury in an attempt toestablish differences between fibrinogenand fibrin. They found that fue two pro-teins exhibited idehtical diffraction pat-terns, which in turn were indistinguish-able from those Astbury had found ear-lier for certain other fibrous proteinssuch as keratin and myosin.

As a result of fuese experiments andothers fibrinogen and fibrin were classi-fied in a group of proteins characterizedby a particular repeat distance. The dis-tance was later shown to be unique to"coiled coils," in which two or moreprotein chains in the forro called analpha helix are twisted together, or su-

92

be 340,000 daltons, which correspondsto a total of about 3,000 amino acids. Itsshape was another matter. The physico-chemical data were as consistent with along, thin asymmetric structure as witha swollen, highly hydrated symmetri-cal one. As it turns out, fue data arealso entirely in harmony with a mole-cule composed of linked nodules.

percoiled. The electron micrographs oíHall and Slayter showed a moleculecomposed oí three aligned glob~les, butthe conhections between fue globulescould not be resolved. That indicatedthey must be lessthan 15 angstrom unitsthick; Carolyn Cohen, who is now atBrandeis University, suggested that fueconnections might be the coiled coils.

In the 1940's and 1950's fibrinogen, aswell as many other commonly availableproteins, was subjected to an onslaughtoí physicochemical studies. Its molecu-lar weight was accurately determined to

fue large fibrinogen molecule into man-ageable components by digesting itwith plasmin. They ended up with a seroí core íragments, fue major ones oíwhich, separated on the basis oí sizeby ion-exchange chromatography, weredesignated íragments A through E.Fragments D and E made up the bulk oíthe recovered mass,and there was abouttwice as much oí D as there was oí E.

These íragments, and another ser pro-duced by digestion with fue enzymetrypsin,.were thoroughly investigated byElemer Mihalyi oí the National Heart,

M ore data carne frorn fue enzyrnaticfragrnentation studies. As early as

1961 investigators at the Pasteur In-stitute in Paris decided to break down

an aggregate of individual fibrin monomers generated by the actionof the enzyme tbrombin on fibrinogen molecules. Strands are from aclot formed in intravenous catheter implanted near a patient's heart.

FIBRIN STRANDS, enlarged about 20,000 diameters, enmesh ared blood cell in an electron micrograph made by Emil O. Bernsteinand Eila Kairinen of the Gillette Research Institute. Each strand is

93

Lung, and Blood lnstitute and his co-workers. Their studies led to two majorconclusions. First, thearchitecture of fi-brinogen makes it susceptible to a kindof attacj( that is independent of fue at-tacking enzyme: fue vulnerability of fuesubstrate rather than the preferentialspecificity of the enzyme is fue directivefeature. This is just what would be ex-pected for a structure made up of linkedglobules. Second, the globules them-selves have a certain amount of rota-tional independence in the intact mole-cule, that is, in solution they are not rig-

idly oriented in space with respect to oneanother. These experiments were fol-lowed by the findings of Victor J. Mar-der of fue National Institute of Arthri-lis and Metabolic Diseases, who studiedfue time course of the plasmin-directedbreakdownof fibrinogen. His character-ization oí transient intermediate break-down products was consistent with asituation in which fue enzyme was suc-cessively cutting links between the do-mains of athree-globule molecule.

Ty¡o other items of structural evi-dence should be mentioned before 1 turn

ba

to the amino acid sequencing. The poly-domainal structure is also supported bya study of fue protein's "melting point."If a protein is heated under carefullyobserved conditions, a temperature isreached at which the protein "melts," ordenatures: fue amino acid chains, for-merly disposed in a precise three-di-mensional conformation, coUapse. Thispoint can be detected by changes in heatflow alone. John W. Donovan and Ron-ald A. Beardslee of fue Western Region-al Research Center of fue U.S. Depart-ment ,oí Agriculture showed that twoproteins in intimate contact melt at asingle temperature that is different fromthe melting point of either of the indi-vidual proteins. Donovan went on toshow, with Mihalyi, that fibrinogen isunique in this respecto It has two meltingpoints, one for the two fragments D andone for fragment E, demonstrating con-vincingly that the two domains are notin intimate contact with each other.

FinaUy, Nancy M. Tooney and Cal ~lyn Cohen studied the structure of ~crocrystals, prepared from partially di-gested fibrinogen, by analyzing optical-diffraction patterns and electron micro-graphs. Their results demonstrated thatthe fundamental unit length of fibrin-ogen must be 450 angstroms; this isin good accord with the 475-angstromlength determined by Hall and Slayterfrom electron microscopy and is alsoconsistent with much of the hydrody-namic data. Recently Cohen has man-aged to obtain genuine crystals fromvariously modified fibrinogens, and al-though the data are stiU at a preliminarystage, they seem to be consistent with amolecule occupying a "cylinder of influ-ence" about 450 angstroms long and 90angstroms in diameter.

c

E stablishing fue amino acid sequenceof a protein as large as fibrinogen ili

a formidable undertaking. It was cle,,)that many aspects of fibrinogen chemis-try could be understood only if sequenceinformation was available, however,and so workers in a number of laborato-ríes began pecking away at the task.

Protein chains have an amino (NH2)end and acarboxy (COOH) end. Whenthe amino acid units at the amino ter-minals of chains in fibrinogen were de-termined, three different "end groups"were identified, two copies of each ofwhich were present in each intact340,000-dalton molecule. Then threedifferent chains were isolated. The totalof their molecular weights carne to onlyhalf fue total weight of the intact mole-cule. These findings made it clear thatthe molecule is composed of two sets ofthree nonidentical chains interconnect-ed by disulfide bonds. (One of fue 20amino acids is cysteine. Two cysteinescan be linked by way of their sulfur at-oms.) The chains were designated alpha,beta and gamma.

FIBRINOGEN AND FIBRIN, negatively stained witb uranyl acetate, are enlarged 375,000diameters (a, b, c) and 500,000 diameters (d) in electron micrograpbs made by Robley C. Wil-liams of the University of California at Berkeley. Individual fibrinogen molecules treated withcalcium iDOS display a tbree-globule structure (a). Thrombin converts fibrinogen joto fibrinmonomers. At an early stage in clot formation (b) two cbains of tbree monomers each are seen;the two chains are offset by a half-monomer overlap. At a later stage (c) a two-chain strandhas developed; the bright regiDos along the strand represent two terminal domains of two mon-omers in end-to-end contacto In a segment of a fully developed, thick fibrin strand (d) the indi-vidual fibrin monomers are oriented horizontally. The prominent vertical striations, 230,000angstrom units apart (half the length of a monomer), are formed by terminal domains in regis-ter tbroughout the three-dimensional strand. The print was made by exposing a negative 10times and offsetting it between exposures by the interval between striations. The techniquereduces "noise" in the micrograph while preserving (and confirming) the periodic structure.

94

TRINODULAR (THREE-GLOBULE) MODEL of the fibrinogen molecule was proposedin 1959 by Cecil E. Hall and Henry S. Slayter of the Massachusetts Institute of Technology.The dimensions are given in angstroms. This early model was based largely on the appear-aRce of the molecule in electron micrographs made by a heavy-metal shadowing technique.

The first major attack on the primarystructure of human fibrinogen was madein Birger Blomback's laboratory at theKarolinska Institute in Stockholm inthe late 1960's. Blomback and his col-leagues treated fibrinogen not with apro tease but with cyanogen bromide,which cleaves protein at only a few sites.They isolated a fragment that containedall the amino terminals of fue three non-identical chains. The fragment was sub-sequently found to be a disulfide-linkeddimer (a compound with two identicalh;ilves), revealing that all six amino ter-minals of fue fibrinogen molecule areclustered near one another in fue intactmolecule.

The fragment accounted for onlyabout 15 percent of the mass of fibrino-gen, but it accoúnted for almost half ofthe two-cysteine units (cystines) consti-tuting a disulfide bond, and so it wasnamed the disulfide knot. Virtually fueOntire sequence of fue fragment, which

ad 247 amino acids ineach of its twohalves, was determined by fue Karolin-ska group. Subsequent experiments byMarder showed that fue disulfide knót isvery similar to fragment E. Both are di-mers, and they cross-react in immuno-logical tests. Moreover, there are twofragments D for every fragment E. Frag-ment E (or the disulfide knot) is fue cen-tral domain of fibrinogen; fue two frag-ments D are fue terminal domains.

After Blomback's initial assault fuesequence offensive stalled for a while,and when fue action resumed, it had.shifted to two other laboratories: AgnesHenschen's at fue Max Planck Institutefor Biochemistry in Munich and oursat fue University of California at SanDiego. After several years of intensivecompetitive effort fue entire covalentstructure was established; most of itwas determined independently in both

O laboratories. The alpha chain consistsof 610 amino acid units, the beta chainof 461 and fue gamma chain of 411, fora total of 1,482 amino acids in eachof fue dimeric halves. The calculatedmolecular weight is 329,842. Approxi-mately 10,000 daltons more is added bytwo different carbohydrate clusters oneach half, bringing fue total into per-fect agreement with fue earlier physico-chemical measurements.

~(10,000)" FRAGMENT'r: 150,000

;¡b;¡;¡¡;¡,

t

'FRAGMENT D, 90,000 tFRAGMENT E, 50,000

t

FRAGMENT D. -

DIGESTION OF FIBRINOGEN with the enzyme plasmin was stndied by Victor J. Marderin 1968. The enzyme gave rise successively to the intermediate prodncts X and Y and to thecore fragments D and E, which are shown here (in terms of the Hall-Slayter model) alongwith their molecular weight. The fragments and the time course of their appearance are con-sistent with the enzymatic cleavage of links between the domains of a three-globnle molecule.

The most interestmg aspect of fue di-sulfides, in my view, was revealed dur-ing fue sequencing of the alpha chain.Blomback had found that each of fuethree nonidentical chains incorporatedan unusual setof cysteines in a sequenceof five amino acids: cysteine-X- Y-Z-cys-teine. We then found a second suchsequence in fue alpha chain, 111 ami-no acid units farther along the chain.Would fue other two chains have similarsecond sequences at fue same relativelocations? Within a few days we had ten-tatively answered fue question with anexcited yes, although it took consider-ably more effort to prove it. All threechains have two cysteine-bounded five-unit sequences separated by 111 amin~

acids (in fue alpha and gamma ch!!ins)or by 112 (in fue beta chain). Two obvi-ous structural questions arose: How arefue disulfide bonds arranged betweenpairs of fuese particular cysteines, andwhat is fue three-dimensional arrange-ment of fue 111 (or 112) amino acidunits separating the two cysteine-bound-ed sequences in each chain?

Even before we began our experi-ments we deduced a solution to thefirstproblem on evolutionary grounds. Wereasoned that in some ancestral fibrino-gen molecule there had been three iden-tical chains (which had since been"variedby mutation of their genes). It is axi-omatic that fue same sequence in thesame environment must have fue same

1 shall consider a few of $e highlightsrevealed by fue sequence studies, par-

ticularly those that have a direct bear-ing on three-dilnensional structure. Anearlynotion that fue three chains mightbe homologous (similar enough to havebeen derived from a common ancestralchain) was increasingly confirmed assequence data accumulated. As in somany proteins, fue homology was bestexemplified by fue locationof fue cyste-ines, all of which, Henschen had longsince found, participate in the forma-tion of disulfide bonds.

95

multiplied by 111, the distance betweendisulfide rings comes to 158.7 ang-stroms. That is in good agreement withfue distance between globules in theHall-Slayter model, which is about 150angstroms.

The only significant regions of alpha-helical structure in any of the chains ap-pear to be those between the disulfiderings. The carboxy-terminal halves offue beta and gamma chains, which arestrongly homologous, appear to be high-ly folded, compact regions; they mateup the bulk of the terminal fragments D.The alpha chain's carboxy half, on fueother hand, is an unfolded, highly po-lar protuberance that is readily cleavedaway from the parent molecule by a va-riety of proteases. Fibrinogen lackingfuese alpha-chain extensions is closelyakin to fue intermediate Marder calledfragment X.

three-dimensional arrangement, and soa structure with a threeíold symmetrymust be evoked, so that each chain isequivalent. This calls íor an assemblagein which disulfide connections link fuefirst cysteine oí one chain to fue secondcysteine oí fue second chain, fue firstcysteine oí the second chain to fue sec-ond cysteine oí fue third chain and final-iy fue first cysteine oí fue third chain tofue second cysteine oí fue first chain,thereby completing a "disulfide ring."When we constructed a detailed modeloí such a structure, the cysteine sidechains íell into place and fue disulfidebonds íormed just such a ring. All fuebiochemical data that have accumulat-ed since then support fue proposed in-terconnection.

The nature oí fue sequences betweenthe disulfide rings was even more inter-esting. Examination oí fue sequencesrevealed that amino acids that havenonpolar (hydrophobic, or water-repel-ling) side chains recur rhythmicallythrough much oí each sequence, suggest-ing a helical arrangement. Computerschemes íor assessing fue iikelihood aparticular sequence will have a particu-lar coníormation revealed that much oíeach inter-ring segment ought to be infue íorm oí an alpha helix. It seemedclear that fuese segments must cor-respond to fue coiled coils. By mak-ing models we could show that the se-quences lend themselves to íorming athree-strand rope in which all the polar(hydrophilic) side chains are turned out-ward into fue surrounding water andmost oí the hydrophobic side chains areturned inward to íorm a nonpolar coreoFinally, we were able to estimate fuedistance from one disulfide ring to fuenext. Whereas fue "translationaP' dis-tance íor a standard alpha helix is ap-proximately 1.5 angstroms per amiGoacid unit, supercoiling shortens fue dis-tance, so that each amino acid advancesabout 1.43 angstroms. If this value is

S O much for structure as such. Howdoes that structure relate to func-

tion: to polymerization and fibrinolysis?It was discovered 30 years ago that thetriggering event for the transformationof fibrinogen into fibrin is fue thrombin-catalyzed removal of two pairs of small,polar peptides (short protein chains), fuefibrinopeptides A and B, which are at fueamino terminals of the alpha and betachains respectively. Each fibrinogenmolecule is thereby converted into a fi-brin monomer, and fue monomers linkup spontaneously to forro fue fibrin net-work. Why do fibrin monomers polym-erize whereas fibrinogen molecules donot? One early view was that fue fibrino-peptides, which ordinarily bear a sub-stantial negative charge, keep fue indi-vidual fibrinogen molecules apart bysimple electrostatic repulsion. In 1952lobo D. Ferry of fue University of Wis-consin at Madison proposed that the re-lease of fue negatively charged fibrino-peptides could allow the monomers toapproach one another, whereupon vari-

ous weak, short-range attractive forcesmight come into play.

Ferry made another significant ob-servation about polymerization. It wasknown that fibrin strands exhibit a char-acteristic banded pattem in electronmicrographs, the repeat distance beingabout 230 angstroms. Since this distanceis roughly half the length of a fibrinmonomer, Ferry suggested that polym-erization must proceed by means of lat-eral pairing, with "partial overlappingresulting in two parallel and end-to-endchains with staggered junctions." Thisview was supported by light-scatteringstudies of fue growing polymer.

In another finding bearing on polym-erization, Bailey and Astbury had foundno difference between fue X-ray-diffrac-tion pattems of fibrinogen and of fibrin,leading them to conclude that "fibrin isno other than an insoluble modificationof fibrinogen without any fundamentalchange in molecular plan." In addition

)fue more recent calorimetric study byDonovan and Mihalyi did not revealany significant difference in the meltingpoints of fibrinogen and fibrin. As wellas reinforcing fue fundamental similari-ty of fibrinogen and fue fibrin monomer,this showed that the intermolecular con-tacts in fibrin are not particularly "coop-erative." They must be few in numberand rather stiff; two fibrin units holdeach other like dancers at arm's lengthrather than nestled cheek to cheek.

It is now possible to reexamine someof the early notions about polymeriza-tion in fue light of the amino acid se-quences, beginning with the electrostat-ic interactions. The human fibrinogenmolecule as a whole has about 26 morenegative charges than positive charges,but fue excess negative charge is notspread equally over fue entire molecule.A disproportionate amount of the ex-cess is on the~entral ~omain. Removal -Jof the two fibrmopeptldes A reduces fueexcess negative charge on the central do-main from minus 8 to minus 1. Whenfue fibrinopeptides B are cleaved away,the central domain actually assumes anet positive charge of plus 5. The coiled-coil regions have approximately fuesame number of positively charged sirlechains as they do negatively chargedones, in keeping with fue idea that theseconnectors are not involved in intermo-lecular interactions but instead remainopen and accessible to eventual attackby plasmin.

r: ~

E ach oí fue terminal domains, how-ever, has a net negative charge oí

minus 4. That is neatly compatible witheach terminal domain's having a com-plementary electrostatic interactionwith a positively charged central do-main during polymerization, giving riseto fue staggered, overlapping aggrega-tion oí fibrin monomers proposed byFerry on the basis oí fue banding patternin electron micrographs.

170,000

SIX-CHAIN STRUCTURE of fibrinogen was established by the early 1970's. Three differ-ent amino acids were found to be linked to amino (NH2) groups, which define ORe end of a pro-tein chain, and there were two copies of each "end group." Breakage of the interchain disulfidebonds (broken lines) isolated three different chains: alpha, beta and gamma, whose total mo-lecular weight, 170,000, accounted for half of the weight of the molecule; there are two copiesof each chain. Many disulfide bonds were found to be in a "disulfide knot" incorporating theamino terminals; it was shown to be largely coincident with fragment E. the central domain.

96

FIBRINOPEPTIDE A":--

26

51

76

101

126

151

176

201

226

251

276

301

326

051376

401

426

451

476

501

526

551

576

601

Ala AlanineArg ArginineAsn AsparagineAsp AspartateCys Cysteine

-Gln Glutamine( Glu Glutamate

Gly GlycineHis HistidinelIe Isoleucine

Leu LeucineLys LysineMet MethioninePhe PhenylalaninePro ProlineSer SerineThr ThreonineTrp TryptophanTyr TyrosineVal Valine

AMINO ACm SEQUENCE OF ALPHA CHAIN, established largely in the author's labora-tory, is given in full; the abbreviations of the 20 amino acids are shown at the left. The coloredarrow shows where thrombin cleaves the chain, releasing the short fibrinopeptide A from theamino end of tbe chain.The darker shade of gray indiéates extent of the "coiled coil" connec-toro The eight cysteines, the amino acids that take part in disulfide bonding, are shown in color.

two molecules thick and accommodatesfue halí-molecule staggered overlap.

Given fue general notion oí interact-ing knobs and holes, with the knobs onthe central domain guarded by fibrino-peptides and fue holes always availableon the terminal domains, how many arethere oí each, and are all fue sets thesame? Aíter all, there are two kinds oífibrinopeptides, A and B. and they areknown to be released at different rates.Torvard C. Laurent and Blomback longago suggested that fue removal oí thefibrinopeptides A. which in mammaliansystems happens first, allows the íor-mation oí polymers that are primarilyderived írom end-to-end interactions,whereas fue subsequent release oí thefibrinopeptides B could allow a lateralinteraction leading to thick fibers. Theiridea was based on early suggestions

What is the mechanism oí fue interac-tion? A simple postulate would be thatthere are specific polymerization siteson the monomers, which are shielded, infibrinogen, by the fibrinopeptides. Thenotion oí specific polymerization sitessuggests complementary structures,which is to say donor sites and acceptorsites, or "knobs" and "holes." For bothefficiency and economy only the knobsor the holes would have to be shieldedby fue fibrinopeptides. Assume that fueholes are always available. In that casefue release oí the fibrinopeptides írom acentral domain could expose a set oíknobs, which could then interact withholes on the terminal domains oí neigh-boring molecules. Since fue startingmonomer is double-ended, fue initialtwo-monomer assembly can grow in ei-ther direction. The growing polymer is

about fiber polymerization and consti-tution advanced by Ferry and PeterR. Morrison, who were both then atHarvard University. Noting fue differ-ence in the opacity of fibrin gels formedunder various conditions, they distin-guished between "fine clots," in whichthe strands seemed to be protofibrilsonly one or two molecules thick, and"coarse clots," with thick fibers made upof bundle s of the protofibrils associatedlaterally. There ought, then, to be twodistinct sets of knobs, one set guarded bythe fibrinopeptides A and one by thefi-brinopeptides B. and two sets of comple-mentary holes.

From fue start it had seemed likelythat fue amino terminals exposed by fuethrombin-catalyzed removal of the fi-brinopeptides were themselves fue con-tact sites for polymerization, and so

97

f ALPHA

C2

FIBRINOPEPTIDE A~ ti e\'\

./ ~,o" " r. V- ~(~

,~ ('" (; "

FIBRINOPEPTIDE B -I",.",,~

~ -~~~~~ (~)E)@E~~--~§~~~

~@@M~CD~oa:1-CD(!)Z<1:o'"

"'"'"

BETA /DISULFIDERINGb

~

~, 'ALPHA~'

, 'BETAGAMMA

COILED-COIL CONNECTOR

f~:j ,,;..~ 230 ANGSTROMS

DETAILED MODEL OF FIBRINOGEN was devised by the au- chains: alpha (gray), befa (white) and gamma (color). The chains arethor and bis colIeagues to take account of the complete amilllo acid linked by 29 disulfide bonds (S-S), 13 within each hall of the dimersequence. The molecule is a dimer; the vertical broken line shows and three connecting the two halves; each "disulfide ring" incorpo- ~its axis of rotational symmetry. It is composed of two seis olE three rates three such bonds, as is shown in detail in the ilIustration at the )

my graduate student Andrew P. Lau-dano and 1 decided to synthesize whatshould be synthetic knobs: peptidescorresponding to fue amino-terminalsegments newly exposed by the releaseof fibrinopeptides. Synthetic glycine-proline-arginine tripeptides (or longerpeptides beginning with that sequence),which correspond to the alpha-chainamino-terminal sequence, were found tobind to fibrinogen and certain prepara-tions of fragment D. They also inhibitedfue polymerization of fibrin monomers;the synthetic knobs filled fue naturallyoccurring holes and preventéd interac-tion with fue natural knobs on fue mon-omers' central domain. On the otherhand, although peptides correspondingto fue beta-chain sequence (glycine-his-tidine-arginine) exposed by the removalof fibrinopeptide B also bóund to fi-brinogen and to fragment D, they didnot prevent the polymerization of fibrin

gamma chain contributes to forming thehole for the alpha knob. Recently Steph-anie Olexa and Andrei Z. Budzynski ofthe Temple University Medical Schooland Hospital have isolated a peptidefrom fue gamma chain's carboxy endthat binds to the central domain; it ap-pears to be a major part of the fabric offue alpha knob's hole.

monomers in fue human system. Com-petition studies showed that two kindsof sites are involved. Evidently the ini-tial polymerization depends exclusivelyon the alpha-chain knobs.

The location of fue holes on fue termi-nal domains has recently been definedmore precisely. Frits Haverkate of fueUniversity of Leiden showed that thenature of fue fragment D generated byplasmin is influenced by calciumions. Intheir absence digestion proceeds muchfurther, generating small fragments Dinwhich fue gamma chain is significantlyshortened at fue carboxy end. Whenweexamined fue binding of the syntheticknobs to fuese two kinds of D, we foundthat fue alpha knob bound only to thelarger D; all binding was abolished whenthe gamma chain was shortened. In con-trast, fue beta knob bound equally wellto both kinds of fragment D. Evident-ly the carboxy-terminal segment of the

T he knob-and-hole links formed un-der fue influence of thrombin alone

involve attractive forces (including elec-trostatic interactions and perhaps al sohydrogen bonding) that can be rathereasily disrupted. Under physiological(asopposed to experimental) conditionsfibrin strands are strengthened, and fuefibrÍn gel is stabilized, by the introduc-tion of covalent bonds between adjacentmonomers. This is accomplished by anenzyme called factor XIII, which is ac-tivated (by the thrombin-catalyzed re-

)

a

ALPHA

/FIRST DISULFIDE RING45 49

SETA

GAMMA

Gln Leu Gln Glu Ala Leu Leu Gln Gln Glu /lrg Pro lIe Arg Asn Ser Val Asp Glu Leu Asn Asn Asn .Val Glu Ala Val Ser Gln Thr Ser Ser Ser Ser Gln Phe Tyr

23 1 ~19GallEO COll

His Gln Leu Tyr lIe Asp Glu Thr Val Asn Ser Asn lIe Pro Thr ~,sn Leu Arg Val Leu Arg Ser lIe Leu Glu Asn Leu Arg Ser Lys lIe Gln Lys Leu Glu Ser Asp Val Ser Ala Gln Me!

1-COILED COIL

AMINO ACID SEQUENCE OF ALL THREE CHAINS is given(a) for a region incorporating the two disulfide rings. (There is oResuch region in each hall of the fibrinogen dimer, of course.) Over

mucb of tbe regiDo tbe cbaios appear to be coiled coils. Tbis would ac-count for a particular X-ray-diffractioo partero disceroed io earlystudies. Tbe coiled coils also accouot for tbe rodlike parts of tbe early

98

~ ~A~MA

~~ THROMBIN...,~/ytIT I ~

4'9 Q/y~~~-t~

-SS~

f'

-

'fA;~~ ~L )~/ ...,

~

-

bottom of these two pages. The connector region between disulfide( -rings is largely composed of supercoiled alpha helixes (coiled coils),

-with a relatively unstructured stretch in the center of each connec-toro In the terminal domains the beta and gamma chains are com-

ly Patrick A. McKee oí fue Duke Uni-versity Medical Center showed thatmultiple cross-links are also established,more slowly, between alpha chains.

The complete structure oí the gam-ma-gamma cross-link was quickly de-termined. Both fue donor (lysine) andthe acceptor (glutamine) units are nearfue carboxy end oí fue gamma chain.The joined molecules are oriented in anantiparallel manner and are bonded bytwo reciprocal cross-links eight aminoacid units aparto The íact that virtuallyall gamma chains become paired in thishighly restrictive geometry indicatesthat each oí fue units in fue fibrin poly-mer has an equivalent orientation. Be-cause fue carboxy-terminal regions oígamma chains are in fue terminal do-mains oí fibrinogen, plasmin digestionoí gamma-linked fibrm gives rise to aíragment-D dimer, as was first noted inMcKee's laboratory.

A gamma-chai~ cross-link connectstwo monomers tf íorm a dimer, butalpha chains can be cross-linked intoextended polymeric arrays. This meanseach alpha chain is linked to at least twoother alpha chains, so that there must beat least two different glutamine accep-tors and two potentiallysine donors onevery alpha chain. The glutamines wereprecisely located during fue course oíour sequence studies. In human fibrin-ogen they are 38 amino acids apart inabout the middle ~í fue alpha chain. Thelysines that participate in alpha-chaincross-linking appear to be in a cluster200 amino acid units nearer the carboxyend oí fue chain. '

moval of a peptide) from a precursorin the plasma and in platelets. The ac-tive enzyme stitches together neighbor-ing monomers in the fibrin polymer byforming a peptide bond (the same bondthat links successive amino acids of aprotein chain) between the side chains ofsuitably disposed lysines (donors) andglutamines (acceptors) and so cross-linking two chains. Covalently stabi-lized fibrin is mechanically strongerthan non-cross-linked fibrin and is alsomore resistant to enzymatic dissolution.It is not, however, morphologically dif-ferent from the non-cross-linked typewhen viewed in electron micrographs;cross-linking does not seem to intro-duce gross structural rearrangement.

My graduate student Renné Chen and1 found that the first cross-links intro-duced in fibrin during the stabilizationprocess are between the gamma chainsof two adjacent molecules. Subsequent-

As 1 noted at fue beginning of this ar-fi ticle, there has to be a mechanismfor dissolving clqts that are no longerneeded, that becqme dislodged into thegeneral circulati9n or that forro at the

e

b Cys23GAMMA ALPHA

1 ' ~ 'Me! Tyr Leu Leu Lys Asp Leu Trp Gln Lys Arg Gln Lys Gln Val Lys Asp Asn Glu Asn Val Val Asn Glu Tyr Ser Ser Glu Leu Glu Lys

.~I161 165 SECONO OISULFIOE RING

ALPHA

BETA

GAMMABETA

--¿1135139

disulfide ring are shown in color. A molecular model of ORe disul-fide ring (b), se en from the amino end of the molecule, shows confor-mation of cysteine-X-Y-Z-cysteine sequences on the three chains.

HaU-Slayter model: the connectors between the globular central andterminal domains of the molecule. Colored arrows indicate whereplasmin is known to cleave the chains. The three bonds within each

99

pactly folded; the alpha chain is hydrophilic and "ftoats" loos~ly inthe aqueous ~nvironment. Two carbohydrate chains (hexagolls) areattached to each half. Specific amino acids are named at the siteswhere release of the fibrinopeptides forms alpha and befa "knobs."

~~wrong time and place. The architectureoí the fibrin arfar is as suited to fuebreakdown oí fibrin as it is to polymeri-zation, ií not more so. The structuralbasis is clear: the accessibility and vul-nerability oí vital connections. The lim-iting oí intermolecular contacts to fuenodular bodies oí fue fibrin monomerslea ves open channels to the interdomain-

al connectors through which proteolyt-ic agents (including plasmin, its precur-sor plasminogen and the activators thatconvert fue precursor into fue active en-zyme) can diffuse. Furthermore, insteadof having to cut all fue way through asolid rodlike molecule, plasmin has onlyto cleave fue slender three-strand con-nectors. As a result dissolution requires

fue breakage of arelatively small num-per oí bonds. It can be supposed that asection oí a fiber is solubilized by fuecleavage oí any two connectors that arewithin, say, lOto 20 fibrin monomersof each other. The system is therebyreduced to fue "intermediate polymer"stage, reversing the process whereby in-termediate polymers forro in advance of

a

b

e

POLYMERIZATION SCHEME proposed b,ere is consistent withelectron-rnicrographic and structural data. TlIle fibrinogen rnolecule(o) is converted joto a fibrin rnonorner (b) by th,~ release of fibrinopep-tides A and B frorn the arnino terrninals of the alpha and beta chainsrespectively. The newly exposed ends of the alpha chains serve as"knobs" that interact with "holes" in the terlminal dornajo, linkingfibrin rnonorners with the observed half-rnonorner overlap (c). Theassernbly can be extended joto a long "interrnediate polyrner" two

molecules thick (d), which is stabilized by the formation of covalentcross-links between adjacent terminal domains. Intermediate poly-mers are then interwoven laterally to form a fully developed fibrinstrand (e), the thick cablelike structure of which can only be suggest-ed in this drawing. The initial polymerization seems to depend spe-cifically on the alpha knobs. The end-to-end covalent cross-linkingis between two gamma chains. Lateral growth may be mediated bythe beta knobs and then strengthened by alpha-chain cross-linking.

loa

;~~.Ir})!

;.,f'"

t1)i

¡;

¡

;,FIBRIN CLOT is stabilized by the cross-linking of two antiparallel gamma-chain regiDos. Re-ciprocal covalent bonds are formed between two Iysine-glutamine pairs near the carboxy ends.

FACTORXIII

THROMBIN

\V-ACTIVATED

FACTORXIII

IIIIII

'VH

\ +NH3

Ca++

/HCH 1 """""" HH/C,¿H

H""" / """H

H/C\N-H

/O=C

\~H"",,/'HH/C

COV ALENT BOND (C-N) of the cross-link is formed by the condensation of lysine andglutamine side chains, with the removal of an ammonia (NHa) group. The reaction is cata-lyzed by the calcium-dependent enzyme called factor XIII, which is activated by thrombin.

the fibrin network on the way to the for-mation of fue gel. Proteolytic digestionthen continues until these intermediatepolymers are reduced to core fragments.

The dissolution scheme underscoresboth fue open nature of fibrin clots andthe extended polydomainal structure offibrinogen and the fibrin monomer. Theplasmin-generated products of fibrindegradation are remarkably similar tofue products of experimental fibrinogendegradation. Obviously the intermolec-ular associations holding the fibrin unitstogether do not shield the units fromproteolytic attack. That is in contrast tomany other biological systems wherethe interaction of two large moleculesaffords one or both of them a degree ofprotection against enzyme-engendereddegradation. The similarity of the diges-tion products al so implies that there isno gross rearrangement of the basicmolecule during polymerization.

Fibrin stabilized by covalent bondingQ lysed at a distinctly lower rate than

nstabilized fibrin, as Johannes Gorm-sen oí fue University of Copenhagen hasshown. Nevertheless, after complete di-gestion by plasmin fue only observeddifference in fue final products is fuepresence of the dimerized "double D"incross-linked preparations. Under lessdrastic conditions, however, GilbertHudry-Clergeon, working in MichelSuscillon's laboratory at fue Universityoí Grenoble, isolated an entityhe called"D2E" from partial plasmin digests oícross-linked fibrin. The complex, whichwas later alBo identified and character-ized by others and whose existence isgenerally accepted, is composed of a"double D" and a fragment E presum-ably írom another monomer with whichthe two D's were linked prior to diges-tion. The complex" D2E"is exactly whatought to result given a scheme oí inter-d ction involving knobs on one domainnd holes on fue other.1 have tried to show how knowledge

oí the amino acid sequence of fibrino-gen bolsters some long-standing notionsabout the protein's three-dimensionalstructure and general behavior. The se-quence data fiesh out a model in whichtwo large terminal domains are connect-ed to a central region by sets of three-strand ropes, giving rise to a trinodularstructure similar to the one electron mi-croscopists proposed some 20 years ago.This extended polydomainal structure isexquisitely suited to a series of consecu-tive operations-polymerization, stabi-liiation and fibrinolysis-that first stopbleeding and then clear away fue clotto prevent blood-vessel blockage. Nowthat fue molecular details of these phe-nomena are understood it may be possi-ble to design reasonableways oí keepingthe system in balance in people whoseblood tends to clot under the wrong cir-cumstances.

~z.~

\~ ~/--f;R.-

o~r1/

101

¡r; \.~..r-

CLOT IS DISSOL VED by the action of plasrnin, sorne of whose principal points of attackare shown (colored lines). Tbe enzyrne has easy access to the coiled-coil connectors, which itcleaves fint in tbe relatively unstructured rniddle region. Plasrnin digestion of cross-linked 6-brin gives rise to "D~» fragrnents stabilized by covalent bonding and knob-to-hole links.