SCIENTIFICAMERICAN

ENZYME MOLECULE

/966

The Three-dimensional Structureof an Enzyme Molecule

The arrangement of atoms in an enzyme molecule has been worked out

for the first time. The enzyme is lysozyme, which breaks open cells

of bacteria. The study has also shown how lysozyme performs its task

by David C. Phillips

O ne day in 1922 Alexander Flem-ing was suffering from a cold.This is not unusual in London,

but Fleming was a most unusual manand he took advantage of the cold in acharacteristic way. He allowed a fewdrops of his nasal mucus to fall on aculture of bacteria he was working withand then put the plate to one side tosee what would happen. Imagine hisexcitement when he discovered sometime later that the bacteria near themucus had dissolved away. For a whilehe thought his ambition of finding auniversal antibiotic had been realized.In a burst of activity he quickly estab-lished that the antibacterial action ofthe mucus was due to the presence init of an enzyme; he called this substancelysozyme because of its capacity to lyse,or dissolve, the bacterial cells. Lyso-zyme was soon discovered in many tis-sues and secretions of the human body,in plants and most plentifully of all inthe white of egg. Unfortunately Flem-ing found that it is not effective againstthe most harmful bacteria. He had towait seven years before a strangelysimilar experiment revealed the exis-tence of a genuinely effective antibi-otic: penicillin.

Nevertheless, Fleming's lysozyme hasproved a more valuable discovery thanhe can have expected when its prop-erties were first established. With it,for example, bacterial anatomists havebeen able to study many details of bac-terial structure [see "Fleming's Lyso-zyme," by Robert F. Acker and S. E.Hartsell; SCIENTIFIC AMERICAN, June,I960]. It has now turned out thatlysozyme is the first enzyme whosethree-dimensional structure has been

78

determined and whose properties areunderstood in atomic detail. Amongthese properties is the way in which theenzyme combines with the substance onwhich it acts—a complex sugar in thewall of the bacterial cell.

Like all enzymes, lysozyme is a pro-tein. Its chemical makeup has beenestablished by Pierre Jolles and hiscolleagues at the University of Parisand by Robert E. Canfield of the Co-lumbia University College of Physiciansand Surgeons. They have found thateach molecule of lysozyme obtainedfrom egg white consists of a singlepolypeptide chain of 129 amino acidsubunits of 20 different kinds. A pep-tide bond is formed when two aminoacids are joined following the removal ofa molecule of water. It is customary tocall the portion of the amino acid in-

corporated into a polypeptide chain aresidue, and each residue has its owncharacteristic side chain. The 129-resi-due lysozyme molecule is cross-linkedin four places by disulfide bridgesformed by the combination of sulfur-containing side chains in different partsof the molecule [see illustration on op-posite page}.

The properties of the molecule cannotbe understood from its chemical con-stitution alone; they depend most criti-cally on what parts of the molecule arebrought close together in the foldedthree-dimensional structure. Some formof microscope is needed to examine thestructure of the molecule. Fortunate-ly one is effectively provided by thetechniques of X-ray crystal-structureanalysis pioneered by Sir LawrenceBragg and his father Sir William Bragg.

ALAARGASNASPCYSGLUGLN

ALANINEARGININEASPARAGINEASPARTIC ACIDCYSTEINEGLUTAMIC ACIDGLUTAMINE

GLYHIS

ILEULEULYSMETPHE

GLYCINEHISTIDINEISOLEUCINELEUCINELYSINEMETHIONINEPHENYLA1 ANINE

PROSERTHRTRYTYRVAL

PROLINESERINETHREONINETRYPTOPHANTYROSINEVALINE

TWO-DIMENSIONAL MODEL of the lysozyme molecule is shown on the opposite page.Lysozyme is a protein containing 129 amino acid subunits, commonly called residues (seekey to abbreviations above). These residues form a polypeptide chain that is cross-linked atfour places by disulfide (-S-S-) bonds. The amino acid sequence of lysozyme was deter-mined independently by Pierre Jolles and his co-workers at the University of Paris and byRobert E. Canfield of the Columbia University College of Physicians and Surgeons. Thethree-dimensional structure of the lysozyme molecule has now been established with thehelp of X-ray crystallography by the author and his colleagues at the Royal Institution inLondon. A painting of the molecule's three-dimensional structure appears on pages 80 and81. The function of lysozyme is to split a particular long-chain molecule, a complex sugar,found in the outer membrane of many living cells. Molecules that are acted on by enzymesare known as substrates. The substrate of lysozyme fits into a cleft, or pocket, formed by thethree-dimensional structure of the lysozyme molecule. In the two-dimensional model onthe opposite page the amino acid residues that line the pocket are shown in dark green.

30

>.

JL

MAIN CHAINCARBON

SIDE CHAINCARBON

THREE-DIMENSIONAL MODEL of the ly-sozyme molecule, painted by Irving Geis, isbased on an actual model assembled at theRoyal Institution by the author and his col-leagues. The painting enables one to traceand distinguish between the chemical bondsthat hold together the main polypeptidechain and the bonds in the 129 side chains,one for each amino acid residue. The mole-cule is folded so as to form a cleft that holdsthe substrate molecule while it is beingbroken in two. The painting on the nextpage shows how the substrate fits into thecleft. The red balls represent oxygen atomsthat are important in splitting the substrate.

NITROGEN

OXYGEN

SULFUR

HYDROGENBOND

\

\

L

The difficulties of examining mole-cules in atomic detail arise, of course,from the fact that molecules are verysmall. Within a molecule each atomis usually separated from its neighborby about 1.5 angstrom units (1.5 X 10'8

centimeter). The lysozyme molecule,which contains some 1,950 atoms, isabout 40 angstroms in its largest di-mension. The first problem is to find amicroscope in which the atoms can beresolved from one another, or seen sep-arately.

The resolving power of a microscopedepends fundamentally on the wave-length of the radiation it employs. Ingeneral no two objects can be seen sep-arately if they are closer together thanabout half this wavelength. The short-est wavelength transmitted by opticalmicroscopes (those working in the ul-traviolet end of the spectrum) is about2,000 times longer than the distancebetween atoms. In order to "see" atomsone must use radiation with a muchshorter wavelength: X rays, which havea wavelength closely comparable tointeratomic distances. The employmentof X rays, however, creates other dif-ficulties: no satisfactory way has yetbeen found to make lenses or mirrorsthat will focus them into an image. Theproblem, then, is the apparently im-possible one of designing an X-raymicroscope without lenses or mirrors.

Consideration of the diffraction the-ory of microscope optics, as developedby Ernst Abbe in the latter part of the19th century, shows that the problemcan be solved. Abbe taught us that theformation of an image in the micro-scope can be regarded as a two-stageprocess. First, the object under exami-nation scatters the light or other radia-tion falling on it in all directions, form-ing a diffraction pattern. This patternarises because the light waves scatteredfrom different parts of the object com-bine so as to produce a wave of largeor small amplitude in any direction

according to whether the waves are inor out of phase—in or out of step—with one another. (This effect is seenmost easily in light waves scatteredby a regularly repeating structure, suchas a diffraction grating made of linesscribed at regular intervals on a glassplate.) In the second stage of imageformation, according to Abbe, the ob-jective lens of the microscope collectsthe diffracted waves and recombinesthem to form an image of the object.Most important, the nature of the im-age depends critically on how much ofthe diffraction pattern is used in itsformation.

X-Ray Structure Analysis

In essence X-ray structure analysismakes use of a microscope in whichthe two stages of image formation havebeen separated. Since the X rays can-not be focused to form an image di-rectly, the diffraction pattern is re-corded and the image is obtained fromit by calculation. Historically the meth-od was not developed on the basis ofthis reasoning, but this way of regard-ing it (which was first suggested byLawrence Bragg) brings out its essen-tial features and also introduces themain difficulty of applying it. In re-cording the intensities of the diffractedwaves, instead of focusing them to forman image, some crucial information islost, namely the phase relations amongthe various diffracted waves. Withoutthis information the image cannot beformed, and some means of recoveringit has to be found. This is the well-known phase problem of X-ray crys-tallography. It is on the solution of theproblem that the utility of the methoddepends.

The term "X-ray crystallography" re-minds us that in practice the methodwas developed (and is still applied) inthe study of single crystals. Crystalssuitable for study may contain some

MODEL OF SUBSTRATE shows how it fits into the cleft in the lysozyme molecule. All thecarbon atoms in the substrate are shown in purple. The portion of the substrate in intimatecontact with the underlying enzyme is a polysaccharide chain consisting of six ringlike struc-tures, each a residue of an amino-sugar molecule. The substrate in the model is made up ofsix identical residues of the amino sugar called N-acetylglucosamine (NAG). In the actualsubstrate every other residue is an amino sugar known as N-acetylmuramic acid (NAM).The illustration is based on X-ray studies of the way the enzyme is bound to a trisaccharidemade of three NAG units, which fills the top of the cleft; the arrangement of NAG units inthe bottom of the cleft was worked out with the aid of three-dimensional models. The sub-strate is held to the enzyme by a complex network of hydrogen bonds. In this style of model-making each straight section of chain represents a bond between atoms. The atoms them-selves lie at the intersections and elbows of the structure. Except for the four red balls rep-resenting oxygen atoms that are active in splitting the polysaccharide substrate, no attemptis made to represent the electron shells of atoms because they would merge into a solid mass.

1015 identical molecules in a regulararray; in effect the molecules in such acrystal diffract the X radiation as thoughthey were a single giant molecule. Thecrystal acts as a three-dimensional dif-fraction grating, so that the waves scat-tered by them are confined to a numberof discrete directions. In order to obtaina three-dimensional image of the struc-ture the intensity of the X rays scatteredin these different directions must bemeasured, the phase problem must besolved somehow and the measurementsmust be combined by a computer.

The recent successes of this methodin the study of protein structures havedepended a great deal on the develop-ment of electronic computers capableof performing the calculations. Theyare due most of all, however, to thediscovery in 1953, by M. F. Perutz ofthe Medical Research Council Labora-tory of Molecular Biology in Cambridge,that the method of "isomorphous re-placement" can be used to solve thephase problem in the study of proteincrystals. The method depends on thepreparation and study of a series of pro-tein crystals into which additional heavyatoms, such as atoms of uranium, havebeen introduced without otherwise af-fecting the crystal structure. The firstsuccesses of this method were in thestudy of sperm-whale myoglobin byJohn C. Kendrew of the Medical Re-search Council Laboratory and in Pe-rutz' own study of horse hemoglobin.For their work the two men received theNobel prize for chemistry in 1962 [see"The Three-dimensional Structure of aProtein Molecule," by John C. Kendrew,SCIENTIFIC AMERICAN, December, 1961,and "The Hemoglobin Molecule," byM. F. Perutz, SCIENTIFIC AMERICAN,November, 1964].

Because the X rays are scattered bythe electrons within the molecules, theimage calculated from the diffractionpattern reveals the distribution of elec-trons within the crystal. The electrondensity is usually calculated at a regu-lar array of points, and the image ismade visible by drawing contour linesthrough points of equal electron den-sity. If these contour maps are drawnon clear plastic sheets, one can obtaina three-dimensional image by assem-bling the maps one above the other in astack. The amount of detail that can beseen in such an image depends on theresolving power of the effective micro-scope, that is, on its "aperture," or theextent of the diffraction pattern that hasbeen included in the formation of theimage. If the waves diffracted throughsufficiently high angles are included

83

(corresponding to a large aperture), theatoms appear as individual peaks inthe image map. At lower resolutiongroups of unresolved atoms appear withcharacteristic shapes by which they canbe recognized.

The three-dimensional structure oflysozyme crystallized from the white ofhen's egg has been determined in atom-ic detail with the X-ray method by ourgroup at the Royal Institution in Lon-

don. This is the laboratory in whichHumphry Davy and Michael Faradaymade their fundamental discoveries dur-ing the 19th century, and in which theX-ray method of structure analysis wasdeveloped between the two world warsby the brilliant group of workers ledby William Bragg, including J. D. Ber-nal, Kathleen Lonsdale, VV. T. Astbury,J. M. Robertson and many others. Ourwork on lysozyme was begun in 1960

when Roberto J. Poljak, a visiting work-er from Argentina, demonstrated thatsuitable crystals containing heavy atomscould be prepared. Since then C. C. F.Blake, A. C. T. North, V. R. Sarma,Ruth Fenn, D. F. Koenig, Louise N.Johnson and G. A. Mair have playedimportant roles in the work.

In 1962 a low-resolution image ofthe structure was obtained that revealedthe general shape of the molecule and

i

LYSOZYME, MAIN CHAIN

LYSOZYME, SIDE CHAIN

SUBSTRATE, MAIN CHAIN

SUBSTRATE, SIDE CHAIN

HYDROGEN BOND

DISULFIDE BOND

102

68

MAP OF LYSOZYME AND SUBSTRATE depicts in color thecentral chain of each molecule. Side chains have been omitted ex-cept for those that produce the four disulfide bonds clipping thelysozyme molecule together and those that supply the terminal con-

nections for hydrogen bonds holding the substrate to the lysozyme.The top three rings of the substrate \A, B,C) are held to the un-derlying enzyme by six principal hydrogen bonds, which are iden-tified by number to key with the description in the text. The lyso-

84

showed that the arrangement of thepolypeptide chain is even more com-plex than it is in myoglobin. This low-resolution image was calculated fromthe amplitudes of about 400 diffractionmaxima measured from native proteincrystals and from crystals containingeach of three different heavy atoms.In 1965, after the development of moreefficient methods of measurement andcomputation, an image was calculated

.118

zyme molecule fulfills its function when itcleaves the substrate between the D and theE ring. Note the distortion of the D ring,which pushes four of its atoms into a plane.

on the basis of nearly 10,000 diffrac-tion maxima, which resolved featuresseparated by two angstroms. Apart fromshowing a few well-separated chlorideions, which are present because thelysozyme is crystallized from a solutioncontaining sodium chloride, the two-angstrom image still does not show in-dividual atoms as separate maxima inthe electron-density map. The level ofresolution is high enough, however, formany of the groups of atoms to be clear-ly recognizable.

The Lysozyme Molecule

The main polypeptide chain appearsas a continuous ribbon of electron den-sity running through the image withregularly spaced promontories on it thatare characteristic of the carbonyl groups(CO) that mark each peptide bond.In some regions the chain is folded inways that are familiar from theoreticalstudies of polypeptide configurationsand from the structure analyses of myo-globin and fibrous proteins such as thekeratin of hair. The amino acid residuesin lysozyme have now been designatedby number; the residues numbered 5through 15, 24 through 34 and 88through 96 form three lengths of "alphahelix," the conformation that was pro-posed by Linus Pauling and Robert B.Corey in 1951 and that was found byKendrew and his colleagues to be themost common arrangement of the chainin myoglobin. The helixes in lysozyme,however, appear to be somewhat dis-torted from the "classical" form, inwhich four atoms (carbon, oxygen, ni-trogen and hydrogen) of each peptidegroup lie in a plane that is parallel tothe axis of the alpha helix. In the lyso-zyme molecule the peptide groups inthe helical sections tend to be rotatedslightly in such a way that their COgroups point outward from the helixaxes and their imino groups (NH) in-ward.

The amount of rotation varies, beingslight in the helix formed by residues 5through 15 and considerable in the oneformed by residues 24 through 34. Theeffect of the rotation is that each NHgroup does not point directly at the COgroup four residues back along the chainbut points instead between the COgroups of the residues three and fourback. When the NH group points di-rectly at the CO group four residuesback, as it does in the classical alphahelix, it forms with the CO group a hy-drogen bond (the weak chemical bondin which a hydrogen atom acts as a

bridge). In the lysozyme helixes the hy-drogen bond is formed somewhere be-tween two CO groups, giving rise to astructure intermediate between that ofan alpha helix and that of a more sym-metrical helix with a three-fold symme-try axis that was discussed by LawrenceBragg, Kendrew and Perutz in 1950.There is a further short length of helix(residues 80 through 85) in which thehydrogen-bonding arrangement is quiteclose to that in the three-fold helix, andalso an isolated turn (residues 119through 122) of three-fold helix. Fur-thermore, the peptide at the far end ofhelix 5 through 15 is in the conforma-tion of the three-fold helix, and the hy-drogen bond from its NH group is madeto the CO three residues back ratherthan four.

Partly because of these irregularitiesin the structure of lysozyme, the pro-portion of its polypeptide chain in thealpha-helix conformation is difficult tocalculate in a meaningful way for com-parison with the estimates obtained byother methods, but it is clearly lessthan half the proportion observed inmyoglobin, in which helical regionsmake up about 75 percent of the chain.The lysozyme molecule does include,however, an example of another regu-lar conformation predicted by Paulingand Corey. This is the "antiparallelpleated sheet," which is believed to bethe basic structure of the fibrous pro-tein silk and in which, as the name sug-gests, two lengths of polypeptide chainrun parallel to each other in oppositedirections. This structure again is stabi-lized by hydrogen bonds between theNH and CO groups of the main chain.Residues 41 through 45 and 50 through54 in the lysozyme molecule form sucha structure, with the connecting resi-dues 46 through 49 folded into a hair-pin bend between the two lengths ofcomparatively extended chain. The re-mainder of the polypeptide chain isfolded in irregular ways that have nosimple short description.

Even though the level of resolutionachieved in our present image was notenough to resolve individual atoms,many of the side chains characteristicof the amino acid residues were readilyidentifiable from their general shape.The four disulfide bridges, for ex-ample, are marked by short rods of highelectron density corresponding to thetwo relatively dense sulfur atoms with-in them. The six tryptophan residuesalso were easily recognized by the ex-tended electron density produced bythe large double-ring structures in their

85

FIRST 56 RESIDUES in lysozyme molecule contain a higher proportion of symmetricallyorganized regions than does all the rest of the molecule. Residues 5 through 15 and 24through 34 (right) form two regions in which hydrogen bonds (gray) hold the residues ina helical configuration close to that of the "classical" alpha helix. Residues 41 through 45and 50 through 54 (left) fold back against each other to form a "pleated sheet," also heldtogether by hydrogen bonds. In addition the hydrogen bond between residues 1 and 40ties the first 40 residues into a compact structure that may have been folded in this waybefore the molecule was fully synthesized (see illustration at the bottom of these two pages).

side chains. Many of the other residuesalso were easily identifiable, but it wasnevertheless most important for therapid and reliable interpretation of theimage that the results of the chemicalanalysis were already available. Withtheir help more than 95 percent of theatoms in the molecule were readilyidentified and located within about .25angstrom.

Further efforts at improving the ac-curacy with which the atoms have beenlocated is in progress, but an almostcomplete description of the lysozymemolecule now exists [see illustration onpages 80 and 81]. By studying it and

GROWING POLYPEPTIDE CHAIN

RIBOSOME

MESSENGER RNA

the results of some further experimentswe can begin to suggest answers to twoimportant questions: How does a mole-cule such as this one attain its observedconformation? How does it function asan enzyme, or biological catalyst?

Inspection of the lysozyme moleculeimmediately suggests two generaliza-tions about its conformation that agreewell with those arrived at earlier in thestudy of myoglobin. It is obvious thatcertain residues with acidic and basicside chains that ionize, or dissociate,on contact with water are all on thesurface of the molecule more or lessreadily accessible to the surrounding

i i i i

liquid. Such "polar" side chains arehydrophilic—attracted to water; theyare found in aspartic acid and glutamicacid residues and in Iysine, arginineand histidine residues, which have basicside groups. On the other hand, mostof the markedly nonpolar and hydro-phobic side chains (for example thosefound in leucine and isoleucine resi-dues) are shielded from the surroundingliquid by more polar parts of the mole-cule. In fact, as was predicted by SirEric Rideal (who was at one time direc-tor of the Royal Institution) and IrvingLangmuir, lysozyme, like myoglobin, isquite well described as an oil drop witha polar coat. Here it is important tonote that the environment of each mole-cule in the crystalline state is not signifi-cantly different from its natural environ-ment in the living cell. The crystalsthemselves include a large proportion(some 35 percent by weight) of mostlywatery liquid of crystallization. Theeffect of the surrounding liquid on theprotein conformation thus is likely to bemuch the same in the crystals as it is insolution.

It appears, then, that the observedconformation is preferred because in itthe hydrophobic side chains are keptout of contact with the surroundingliquid whereas the polar side chainsare generally exposed to it. In this waythe system consisting of the proteinand the solvent attains a minimum freeenergy, partly because of the large num-ber of favorable interactions of likegroups within the protein molecule andbetween it and the surrounding liquid,and partly because of the relativelyhigh disorder of the water moleculesthat are in contact only with otherpolar groups of atoms.

Guided by these generalizations,many workers are now interested in thepossibility of predicting the conforma-

CODON NUMBER 1 10J L

FOLDING OF PROTEIN MOLECULE may take place as the grow-ing polypeptide chain is being synthesized by the intracellular par-ticles called ribosomes. The genetic message specifying the amino

acid sequence of each protein is coded in "messenger" ribonucleicacid (RNA). It is believed several ribosomes travel simultaneouslyalong this long-chain molecule, reading the message as they go.

tion of a protein molecule from itschemical formula alone [see "MolecularModel-building by Computer," by Cy-rus Levinthal; SCIENTIFIC AMERICAN,June]. The task of exploring all possibleconformations in the search for the oneof lowest free energy seems likely, how-ever, to remain beyond the power of anyimaginable computer. On a conservativeestimate it would be necessary to con-sider some 1011>0 different conformationsfor the lysozyme molecule in any gen-eral search for the one with minimumfree energy. Since this number is fargreater than the number of particles inthe observable universe, it is clear thatsimplifying assumptions will have to bemade if calculations of this kind are tosucceed.

The Folding of Lysozyme

For some time Peter Dunnill and Ihave been trying to develop a model ofprotein-folding that promises to makepracticable calculations of the minimumenergy conformation and that is, at thesame time, qualitatively consistent withthe observed structure of myoglobinand lysozyme. This model makes useof our present knowledge of the wayin which proteins are synthesized inthe living cell. For example, it is wellknown, from experiments by Howard M.Dintzis and by Christian B. Anfinsenand Robert Canfield, that protein mole-cules are synthesized from the terminalamino end of their polypeptide chain.The nature of the synthetic mecha-nism, which involves the intracellular

. particles called ribosomes working incollaboration with two forms of ribonu-cleic acid ("messenger" RNA and "trans-fer" RNA), is increasingly well under-stood in principle, although the detailedenvironment of the growing proteinchain remains unknown. Nevertheless,

it seems a reasonable assumption that,as the synthesis proceeds, the aminoend of the chain becomes separated byan increasing distance from the point ofattachment to the ribosome, and that thefolding of the protein chain to its na-tive conformation begins at this endeven before the synthesis is complete.According to our present ideas, partsof the polypeptide chain, particularlythose near the terminal amino end, mayfold into stable conformations that canstill be recognized in the finished mole-cule and that act as "internal templates,"or centers, around which the rest ofthe chain is folded [see illustration atbottom of these two pages]. It maytherefore be useful to look for the stableconformations of parts of the polypep-tide chain and to avoid studying all thepossible conformations of the wholemolecule.

Inspection of the lysozyme moleculeprovides qualitative support for theseideas [see top illustration on oppositepage]. The first 40 residues from theterminal amino end form a compactstructure (residues 1 and 40 are linkedby a hydrogen bond) with a hydropho-bic interior and a relatively hydrophilicsurface that seems likely to have beenfolded in this way, or in a simply re-lated way, before the molecule wasfully synthesized. It may also be im-portant to observe that this part ofthe molecule includes more alpha helixthan the remainder does.

These first 40 residues include amixture of hydrophobic and hydro-philic side chains, but the next 14 resi-dues in the sequence are all hydrophilic;it is interesting, and possibly significant,that these are the residues in the anti-parallel pleated sheet, which lies outof contact with the globular submole-cule formed by the earlier residues. Inthe light of our model of protein fold-

ing the obvious speculation is thatthere is no incentive to fold these hy-drophilic residues in contact with thefirst part of the chain until the hy-drophobic residues 55 (isoleucine) and56 (leucine) have to be shielded fromcontact with the surrounding liquid.It seems reasonable to suppose thatat this stage residues 41 through 54fold back on themselves, forming thepleated-sheet structure and burying thehydrophobic side chains in the initialhydrophobic pocket.

Similar considerations appear to gov-ern the folding of the rest of the mole-cule. In brief, residues 57 through 86are folded in contact with the pleated-sheet structure so that at this stageof the process—if indeed it follows thiscourse—the folded chain forms a struc-ture with two wings lying at an angleto each other. Residues 86 through 96form a length of alpha helix, one sideof which is predominantly hydrophobic,because of an appropriate alternationof polar and nonpolar residues in thatpart of the sequence. This helix lies inthe gap between the two wings formedby the earlier residues, with its hydro-phobic side buried within the molecule.The gap between the two wings is notcompletely filled by the helix, however;it is transformed into a deep cleft run-ning up one side of the molecule. As weshall see, this cleft forms the active siteof the enzyme. The remaining residuesare folded around the globular unitformed by the terminal amino end ofthe polypeptide chain.

This model of protein-folding can betested in a number of ways, for exampleby studying the conformation of thefirst 40 residues in isolation both di-

Presumably the messenger RNA for lysozyme contains 129 "co-dons," one for each amino acid. Amino acids are delivered to thesite of synthesis by molecules of "transfer" RNA (dark color). The

illustration shows how the lysozyme chain would lengthen as a ri-bosome travels along the messenger RNA molecule. Here, hypothet-ically, the polypeptide is shown folding directly into its final shape.

87

rectly (after removal of the rest of themolecule) and by computation. Ulti-mately, of course, the model will be re-garded as satisfactory only if it helps usto predict how other protein moleculesare folded from a knowledge of theirchemical structure alone.

The Activity of Lysozyme

In order to understand how lysozymebrings about the dissolution of bacte-ria we must consider the structure ofthe bacterial cell wall in some detail.Through the pioneer and independentstudies of Karl Meyer and E. B. Chain,followed up by M. R. J. Salton of theUniversity of Manchester and many oth-ers, the structures of bacterial cell wallsand the effect of lysozyme on them arenow quite well known. The importantpart of the cell wall, as far as lysozymeis concerned, is made up of glucose-likeamino-sugar molecules linked togetherinto long polysaccharide chains, whichare themselves cross-connected by shortlengths of polypeptide chain. This partof each cell wall probably forms oneenormous molecule—a "bag-shaped mac-romolecule," as W. Weidel and H. Pel-zer have called it.

The amino-sugar molecules concernedin these polysaccharide structures are oftwo kinds; each contains an acetamido(-NH • CO • CH;i) side group, but oneof them contains an additional majorgroup, a lactyl side chain [see illustra-tion below]. One of these ammo sugarsis known as N-acetylglucosamine (NAG)and the other as N-acetylmuramic acid(NAM). They occur alternately in the

polysaccharide chains, being connectedby bridges that include an oxygen atom(glycosidic linkages) between carbonatoms 1 and 4 of consecutive sugarrings; this is the same linkage that joinsglucose residues in cellulose. The poly-peptide chains that cross-connect thesepolysaceharides are attached to theNAM residues through the lactyl sidechain attached to carbon atom 3 in eachNAM ring.

Lysozyme has been shown to breakthe linkages in which carbon 1 in NAMis linked to carbon 4 in NAG but not theother linkages. It has also been shownto break down ehitin, another commonnatural polysaccharide that is found inlobster shell and that contains onlyNAG.

Ever since the work of Svante Ar-rhenius of Sweden in the late 19th cen-tury enzymes have been thought towork by forming intermediate com-pounds with their substrates: the sub-stances whose chemical reactions theycatalyze. A proper theory of the en-zyme-substrate complex, which under-lies all present thinking about enzymeactivity, was clearly propounded byLeonor Michaelis and Maude Men tonin a remarkable paper published in1913. The idea, in its simplest form,is that an enzyme molecule provides asite on its surface to which its substratemolecule can bind in a quite preciseway. Reactive groups of atoms in theenzyme then promote the requiredchemical reaction in the substrate. Ourimmediate objective, therefore, was tofind the structure of a reactive complexbetween lysozyme and its polysaccha-

ride substrate, in the hope that wewould then be able to recognize the ac-tive groups of atoms in the enzyme andunderstand how they function.

Our studies began with the observa-tion by Martin Wenzel and his col-leagues at the Free University of Berlinthat the enzyme is prevented from func-tioning by the presence of NAG itself.This small molecule acts as a competi-tive inhibitor of the enzyme's activityand, since it is a part of the large sub-strate molecule normally acted on bythe enzyme, it seems likely to do this bybinding to the enzyme in the way thatpart of the substrate does. It preventsthe enzyme from working by preventingthe substrate from binding to the en-zyme. Other simple amino-sugar mole-cules, including the trisaccharide madeof three NAG units, behave in the sameway. We therefore decided to study thebinding of these sugar molecules to thelysozyme molecules in our crystals inthe hope of learning something aboutthe structure of the enzyme-substratecomplex itself.

My colleague Louise Johnson soonfound that crystals containing the sugarmolecules bound to lysozyme can beprepared very simply by adding thesugar to the solution from which thelysozyme crystals have been grown andin which they are kept suspended. Thesmall molecules diffuse into the proteincrystals along the channels filled withwater that run through the crystals. For-tunately the resulting change in thecrystal structure can be studied quitesimply. A useful image of the electron-density changes can be calculated from

R,

4 C

IC6

\C 1

• o

R,

HI

— C —OHIH

POLYSACCHARIDE MOLECULE found in the walls of certainbacterial cells is the substrate broken by the lysozyme molecule.The polysaccharide consists of alternating residues of two kinds ofamino sugar: N-acetylglucosamine I NAG I and N-acetylmuramicacid (NAM I. In the length of polysaccharide chain shown here

O HI

—N—C—C—HI IH H

R,

HI

O

— C — C-OHI

H —C — HIH

A, C and £ are NAG residues; li, D and F are NAM residues. Theinset at left shows the numbering scheme for identifying theprincipal atoms in each sugar ring. Six rings of the polysaccharidefit into the cleft of the lysozyme molecule, which effects a cleav-age between rings D and E (see illustration on pages 84 and 85).

88

measurements of the changes in ampli-tude of the diffracted waves, on the as-sunrption that their phase relations havenot changed from those determined forthe pure protein crystals. The imageshows the difference in electron densitybetween crystals that contain the addedsugar molecules and those that do not.

In this way the binding to lysozymeof eight different amino sugars wasstudied at low resolution (that is,through the measurement of changes inthe amplitude of 400 diffracted waves).The results showed that the sugars bindto lysozyme at a number of differentplaces in the cleft of the enzyme. Theinvestigation was hurried on to higherresolution in an attempt to discover theexact nature of the binding. Happilythese studies at two-angstrom resolution(which required the measurement of10,000 diffracted waves) have nowshown in detail how the trisaccharidemade of three NAG units is bound tothe enzyme.

The trisaccharide fills the top half ofthe cleft and is bound to the enzyme bya number of interactions, which can befollowed with the help of the illustra-tion on pages 84 and 85. In this illus-tration six important hydrogen bonds,to be described presently, are identifiedby number. The most critical of theseinteractions appear to involve the aceta-mido group of sugar residue C [thirdfrom top], whose carbon atom 1 is notlinked to another sugar residue. Thereare hydrogen bonds from the CO groupof this side chain to the main-chain NHgroup of amino acid residue 59 in theenzyme molecule [bond No. 1] andfrom its NH group to the main-chainCO group of residue 107 (alanine) inthe enzyme molecule [bond No. 2]. Itsterminal CH3 group makes contact withthe side chain of residue 108 (trypto-phan). Hydrogen bonds [No. 3 and, No.4] are also formed between two oxygenatoms adjacent to carbon atoms 6 and 3of sugar residue C and the side chainsof residues 62 and 63 (both tryptophan)respectively. Another hydrogen bond[No. 5] is formed between the acetami-do side chain of sugar residue A andresidue 101 (aspartic acid) in the en-zyme molecule. From residue 101 thereis a hydrogen bond [No. 6] to the oxy-gen adjacent to carbon atom 6 of sugarresidue B. These polar interactions aresupplemented by a large number ofnonpolar interactions that are more dif-licult to summarize briefly. Among themore important nonpolar interactions,however, are those between sugar resi-due B and the ring system of residue

62; these deserve special mention be-cause they are affected by a smallchange in the conformation of the en-zyme molecule that occurs when thetrisaccharide is bound to it. The elec-tron-density map showing the change inelectron density when tri-NAG is boundin the protein crystal reveals clearlythat parts of the enzyme molecule havemoved with respect to one another.These changes in conformation arelargely restricted to the part of theenzyme structure to the left of the cleft,which appears to tilt more or less as awhole in such a way as to close the cleftslightly. As a result the side chain ofresidue 62 moves about .75 angstromtoward the position of sugar residue B.Such changes in enzyme conformationhave been discussed for some time, no-tably by Daniel E. Koshland, Jr., of theUniversity of California at Berkeley,whose "induced fit" theory of the en-zyme-substrate interaction is supportedin some degree by this observation inlysozyme.

The Enzyme-Substrate Complex

At this stage in the investigation ex-citement grew high. Could we tell howthe enzyme works? I believe we can.Unfortunately, however, we cannot seethis dynamic process in our X-ray im-ages. We have to work out what musthappen from our static pictures. Firstof all it is clear that the complex formedby tri-NAG and the enzyme is not theenzyme-substrate complex involved incatalysis because it is stable. At low con-centrations tri-NAG is known to behaveas an inhibitor rather than as a substratethat is broken down; clearly we havebeen looking at the way in which itbinds as an inhibitor. It is noticeable,however, that tri-NAG fills only half ofthe cleft. The possibility emerges thatmore sugar residues, filling the remain-der of the cleft, are required for theformation of a reactive enzyme-substratecomplex. The assumption here is thatthe observed binding of tri-NAG as aninhibitor involves interactions with theenzyme molecule that also play a partin the formation of the functioning en-zyme-substrate complex.

Accordingly we have built a modelthat shows that another three sugarresidues can be added to the tri-NAGin such a way that there are satisfactoryinteractions of the atoms in the proposedsubstrate and the enzyme. There is onlyone difficulty: carbon atom 6 and itsadjacent oxygen atom in sugar residueD make uncomfortably close contacts

5 OXYGEN

"CHAIR" CONFIGURATION (gray) is thatnormally assumed by the rings of aminosugar in the polysaccharide substrate. Whenbound against the lysozyme, however, the Dring is distorted {color) so that carbonatoms 1, 2 and 5 and oxygen atom 5 lie ina plane. The distortion evidently assists inbreaking the substrate below the D ring.

with atoms in the enzyme molecule, un-less this sugar residue is distorted alittle out of its most stable "chair" con-formation into a conformation in whichcarbon atoms 1, 2 and 5 and oxygenatom 5 all lie in a plane [see illustrationabove]. Otherwise satisfactory interac-tions immediately suggest themselves,and the model falls into place.

At this point it seemed reasonableto assume that the model shows thestructure of the functioning complex be-tween the enzyme and a hexasaccharide.The next problem was to decide whichof the five glycosidic linkages would bebroken under the influence of the en-zyme. Fortunately evidence was at handto suggest the answer. As we have seen,the cell-wall polysaccharide includes al-ternate sugar residues of two kinds,NAG and NAM, and the bond brokenis between NAM and NAG. It wastherefore important to decide which ofthe six sugar residues in our model couldbe NAM, which is the same as NAGexcept for the lactyl side chain append-ed to carbon atom 3. The answer wasclear-cut. Sugar residue C cannot beNAM because there is no room for thisadditional group of atoms. Therefore thebond broken must be between sugarresidues B and C or D and E. We al-ready knew that the glycosidic linkagebetween residues B and C is stablewhen tri-NAG is bound. The conclu-sion was inescapable: the linkage thatmust be broken is the one betweensugar residues D and E.

Now it was possible to search for theorigin of the catalytic activity in theneighborhood of this linkage. Our taskwas made easier by the fact that John A.

89

Rupley of the University of Arizona hudshown that the chemical bond brokenunder the influence of lysozyme is theone between carbon atom 1 and oxygenin the glycosidic link rather than thelink between oxygen and carbon atom4. The most reactive-looking group ofatoms in the vicinity of this bond arethe side chains of residue 52 (asparticacid) and residue 35 (glutamic acid).

One of the oxygen atoms of residue 52is about three angstroms from carbonatom 1 of sugar residue D as well asfrom the ring oxygen atom 5 of thatresidue. Residue 35, on the other hand,is about three angstroms from the oxy-gen in the glycosidic linkage. Further-more, these two amino acid residueshave markedly different environments.Residue 52 has a number of polarneighbors and appears to be involved ina network of hydrogen bonds linking itwith residues 46 and 59 (both aspara-gine) and, through them, with residue50 (serine). In this environment residue52 seems likely to give up a terminalhydrogen atom and thus be negativelycharged under most conditions, evenwhen it is in a markedly acid solution,whereas residue 35, situated in a non-polar environment, is likely to retain itsterminal hydrogen atom.

A little reflection suggests that theconcerted influence of these two amino

^ P CARBON

l |§ | OXYGEN

| j fe HYDROGEN

0 H - _- ' ' WATERMOLECULE

LYSOZYME, MAIN CHAIN

LYSOZYME,MAIN CHAIN

SPLITTING OF SUBSTRATE BY LYSOZYME is believed to involve the proximity andactivity of two side chains, residue 35 (glutaniicacidi and residue 52 (aspartic acid ). It is pro-posed that a hydrogen ion (H+ ) becomes detached from the OH group of residue 35 andattaches itself to the oxygen atom that joins rings D and E, thus breaking the bond betweenthe two rings. This leaves carbon atom 1 of the D ring with a positive charge, in which formit is known as a carbonium ion. It is stabilized in this condition by the negatively chargedside chain of residue 52. The surrounding water supplies an OH~ ion to combine with thecarbonium ion and an H + ion to replace the one lost by residue 35. The two parts ofthe substrate then fall away, leaving the enzyme free to cleave another polysaccharide chain.

acid residues, together with a contribu-tion from the distortion to sugar residueD that has already been mentioned, isenough to explain the catalytic activityof lysozyme. The events leading to therupture of a bacterial cell wall probablytake the following course [see illustra-tion on this page].

First, a lysozyme molecule attachesitself to the bacterial cell wall by in-teracting with six exposed ammo-sugarresidues. In the process sugar residue Dis somewhat distorted from its usualconformation.

Second, residue 35 transfers its ter-minal hydrogen atom in the form of ahydrogen ion to the glycosidic oxygen,thus bringing about cleavage of thebond between that oxygen and carbonatom 1 of sugar residue D. This createsa positively charged carbonium ion (C +)where the oxygen has been severedfrom carbon atom 1.

Third, this carbonium ion is stabilizedby its interaction with the negativelycharged aspartic acid side chain of resi-due 52 until it can combine with a hy-droxyl ion (OH~) that happens to dif-fuse into position from the surroundingwater, thereby completing the reaction.The lysozyme molecule then falls away,leaving behind a punctured bacterialcell wall.

It is not clear from this descriptionthat the distortion of sugar residue Dplays any part in the reaction, but infact it probably does so for a very in-teresting reason. R. H. Lemieux andG. Huber of the National ResearchCouncil of Canada showed in 1955 thatwhen a sugar molecule such as NAG in-corporates a carbonium ion at the car-bon-1 position, it tends to take up thesame conformation that is forced onring D by its interaction with the en-zyme molecule. This seems to be anexample, therefore, of activation of thesubstrate by distortion, which has longbeen a favorite idea of enzymologists.The binding of the substrate to the en-zyme itself favors the formation of thecarbonium ion in ring D that seems toplay an important jsart in the reaction.

It will be clear from this account thatalthough lysozyme has not been seenin action, we have succeeded in build-ing up a detailed picture of how itmay work. There is already a great dealof chemical evidence in agreement withthis picture, and as the result of all thework now in progress we can be surethat the activity of Fleming's lysozymewill soon be fully understood. Best ofall, it is clear that methods now existfor uncovering the secrets of enzymeaction.

90