50 Years of Protein Structure Analysis

50 Years of Protein Structure Analysis

Bror Strandberg1, Richard E. Dickerson2⁎ and Michael G. Rossmann3

Fifty years ago, Max Perutz and John Kendrew at Cambridge Universityachieved something that many people at the time considered impossible:they were the first to use x-ray crystallography to decipher the molecularstructures of proteins: haemoglobin and myoglobin. They found that bothmolecules were built from Linus Pauling's alpha helices, but folded andpacked together in a complicated manner that never could have beendeciphered by any other technique. With structure information in hand theycould then explain how haemoglobin in the bloodstream binds and releasesoxygen on cue, how it passes its cargo on to the related storage proteinmyoglobin, and how a single amino acid mutation can produce the cata-strophe known as sickle-cell anemia. Perutz and Kendrew also observedthat the folding of helices was identical in myoglobin and the two chains ofhaemoglobin, and this along with the simultaneously evolving new tech-nique of amino acid sequence analysis established for the first time theconcept of molecular evolution.

The crystallographic puzzle was qcrackedq by Perutz when he demon-strated that the binding of only two heavy metal atoms to horsehaemoglobin changed the x-ray pattern enough to allow him to solve theqphase problemq and circumvent the main obstacle to protein crystalstructure analysis. Because myoglobin has a single chain whereas haemo-globin has four, Kendrew's work with myoglobin progressed more rapidly;a low resolution structure appeared in 1956 and the high resolution struc-ture in 1959. That same year saw the low resolution picture of haemoglobin,and the high resolution structure followed shortly thereafter.

Much of the work in structure analysis was carried out by visitingpostdoctoral fellows and technicians, under the watchful eye of Perutz andKendrew. This celebratory review has been written by three of those formerpostdoctorals: Strandberg and Dickerson from the myoglobin project, andRossmann from the haemoglobin.

1Department of Cell andMolecular Biology,Uppsala University, Box 596,SE-751 24 Uppsala, SwedenE-mail address:[email protected]

2Molecular Biology Institute,Boyer Hall, University ofCalifornia at Los Angeles,Los Angeles, CA 90095-1570,USAE-mail address:[email protected]

3Hockmeyer Hall of StructuralBiology, Purdue University,249 South Martin Jischke Drive,West Lafayette, IN 47907-1971,USAE-mail address: [email protected]

Received 7 May 2009;accepted 8 May 2009

doi:10.1016/j.jmb.2009.05.024 J. Mol. Biol. (2009) 392, 2–32

Available online at www.sciencedirect.com

Chapter 1: Building the Ground for theFirst Two Protein Structures: Myoglobinand Haemoglobin

Bror Strandberg

Introduction

Just 50 years ago, two Cambridge scientists, MaxPerutz and John Kendrew, achieved a goal that hadlong been considered completely impossible: theysolved the molecular structures of two related pro-teins—myoglobin and haemoglobin—by X-ray crys-tallography. In these days of automatic diffracto-meters; large high-speed computers; sophisticated

*Corresponding author.

methods for purifying, crystallizing, and labelingproteins with heavy atoms; and elegant computerprograms for model building of protein structures, itis easy to forget how challenging the task was in theabsence of all the aforementioned techniques. JamesWatson, who coparticipated a few years earlier in thestructure analysis of DNA, once remarked sardoni-cally that, “In some circles, an interest in the historyof one's field is regarded as a sign of decliningpowers.”ButGeorge Santayanawarns us that, “Thosewho cannot remember the past are condemned to

mailto:[email protected]

http://www.davegranlund.com/cartoons

mailto:[email protected]


repeat it.” So on the 50th year anniversary of such aremarkable achievement, it isworthwhile taking a fewmoments to consider the first time that anyone dis-covered what a protein molecule actually looked like.Over the years, Kendrew and Perutz were assisted

by many young collaborators from different parts ofthe world. Three of them were the authors of thismemorial article. We were very privileged, since wehad the good fortune to take part in the final stagesof the structure determinations and could thereforeexperience the great moments of discovery when, forthe first time, it could be seen how protein moleculeswere constructed. In the following two chapters, mycolleagues Richard E. Dickerson and Michael G.Rossmann describe what happened when the struc-tures of myoglobin and haemoglobin were deter-mined at 2.0 Å and 5.5 Å resolution, respectively.

The First Steps

For a full appreciation of how these importantresults could be obtained, it is useful to have a briefbackground on how the basic method—X-ray crys-tallography—was invented and developed, in par-ticular so that it could be used to determine thestructures of such huge macromolecules. A handfulof people played key roles in this development, andto follow their main discoveries is, of course, ofspecial interest.After the discovery of X-ray radiation by Wilhelm

Röntgen in 1895, it was clear that this new radiationtechnique could have important medical applica-tions. The use of X-rays in other fields was not soobvious. However, this situation changed through adiscovery by the German physicist Max von Laueand his colleagues.The phenomenon of interference was well known

in optics using ordinary light. Light waves hittingregularly spaced objects (e.g., equally spacedscratches in a glass plate) enhance each other incertain directions and work against each other inother directions (here more or less entirely reducingeach other's effect. In this manner, an interferencepattern or diffraction pattern is produced. To obtaininterference, the size of the wavelength has to beof the same order of magnitude as the distancebetween the interfering objects.Since binding distances between atoms vary bet-

ween about 1 Å and 3 Å and since X-ray wave-lengths are on the order of 0.05–100 Å, there aremany X-ray wavelengths that are of the same orderas the binding distances between atoms. This factled von Laue to predict that a crystal, which is aregular repetition of molecules and atoms in space,ought to function as a three-dimensional interfer-ence lattice for X-ray radiation. This prediction wasverified experimentally by von Laue and hiscolleagues Friedrich and Knipping when theyobtained the first X-ray diffraction photographfrom a crystal, that of zinc blende ZnS, in 1912.1

The atoms (or, more correctly, electrons) in thecrystals produce the diffraction of X-rays. Therefore,

the next important development was to use theinformation in X-ray photographs to determine thestructures of compounds built up by the atoms. Thepioneer of this work was a very young scientist,W. L. Bragg, collaborating in part with his father,W. H. Bragg.W. Lawrence Bragg was born in 1890 in Adelaide,

Australia, where he got his basic education. Hisfather, W. H. Bragg, was professor in physics thereuntil 1909, when he was appointed as professor inphysics at the University of Leeds. Lawrence Braggthen started as an undergraduate in mathematicsand physics at Cambridge and had just graduatedwhen the work by von Laue and his colleagues waspublished.What Lawrence Bragg did was to calculate the

intensities of all reflections (interference points) in anX-ray diffraction photograph and to compare thesevalues with the observed intensities. The shortdescription below will suggest how these calcula-tions could be performed.Usually a crystal contains several atoms of

different kinds in the smallest volume (the unitcell), which is translated in three dimensions to buildup the crystal. If there are n atoms in the unit cell, wecan describe the crystal as being constructed by ncongruent three-dimensional lattices inserted intoone another. When X-ray radiation strikes thecrystal, we get diffraction from all lattices in thesame directions, but since the lattices are shiftedrelative to each other, the diffracted waves comingfrom the different atoms in the unit cell are not inphase with one another. If the atom n has thecoordinates xn, yn, zn (values in the range 0–1), thephase difference (the phase angle) for the nth latticerelative to the origin is:

an = 2k hxn + kyn + lznð Þ

for the diffraction order h,k,l (which are the coor-dinates in diffraction space).Thus, when calculating the sum of contributions

from different atoms to every diffraction order(reflection), we have to sum up the waves withdifferent starting points and different amplitudes fn(proportional to the number of electrons in atom n).The result of this summation is called the structurefactor F(hkl) for reflection h,k,l, and its expressionbelow can easily be deduced:

FðhklÞ =X

n

fnd ei2kðhxnþkynþlznÞ =X

n

fneian

The summation can be shown graphically to be asum in the complex plane, and this is viewed forn=3 in Fig. 2.1b of Chapter 2 of this work. The heavyarrow in Fig. 2.1b is the structure factor, whichrepresents the scattering for that reflection from theentire n=3 structure. Figure 2.1c shows the anglebetween the structure factor and the horizontal orreal axis in the complex plane. This angle α(hkl) is akey parameter in structure determination by X-raycrystallography, as we will see further on.

4 50 Years of Protein Structure Analysis

When the atoms are correctly placed in the crys-tal unit cell, the calculated value of the structurefactor |F(hkl)|calc should compare well with theobserved value |F(hkl)|obs, which can be obtainedfrom the formula |F(hkl)|obs=k√I, where k is aconstant and I is the intensity of the diffracted (orscattered) X-ray beam for reflection hkl.Lawrence Bragg calculated structure factors and

compared these with the corresponding observedvalues for several compounds such as sodium chlo-ride (NaCl), potassium chloride (KCl), fluorspar(CaF), and calcite (CaCO3).

2 From the starting posi-tions of the atoms (based mainly on packing consi-derations), the atoms were moved until the calcu-lated and observed structure factors were in the bestpossible agreement. This “trial-and-error” methodfor structure determination worked because thecompounds were simple and the calculated structurefactors were very sensitive to the positions of theatoms.In 1915, W. L. and W. H. Bragg shared the Nobel

Prize in Physics—W. L. Bragg (Fig. 1.1), then only25 years old, for his work on diffraction and crystalstructures, and his father, W. H. Bragg, for studies onthe origin and properties of X-rays.Lawrence Bragg continued to work on the struc-

ture of minerals, but the problems increased whenthe compounds became larger. In his work on thesilicate mineral diopside CaMg(SiO3)2, he had greatdifficulties finding the positions, especially of thelight oxygen atoms. He finally found a solution,3 butasked himself an important question: “Is there aneasier way?” Yes there was, and help came from

Fig. 1.1. W. Lawrence Bragg photographed in 1913when he was 23 years old. Reprinted with permissionfrom The Royal Society, London.

work performed by the mathematician Fourier, whoin the 1850s had already shown that a reasonablybehaving periodic function could be expressed as aseries of exponential terms. If such a function wascombinedwith the expression for the structure factorgiven above, it was possible to obtain values for theelectron density (a well-behaving function) at anypoint x,y,z in the crystal unit cell. Below, I show theexpression deduced for the electron density, mainlyto point out to the reader the two important para-meters needed: the observed structure factor and thephase angle:

U xyzð Þ = 1=V dXXXþl

h;k;l = −l

jFðhklÞjeiaðhklÞd e�i2kðhxþkyþlzÞ

whereV is the volume of the unit cell; |F(hkl)| is theobserved structure factor; h,k,l are the diffractionorders; and α(hkl) is just the phase angle as describedabove and shown in Fig. 2.1c of Chapter 2.If the crystal does not contain any atom that

scatters in an anomalous way (and this is valid inmost cases), then the expression above can be for-mulated as follows:

U xyzð Þ = 2=V dXþl

h = 0

Xþl

k = �l

Xþl

l = �l

jFðhklÞj

d cos 2k hxþkyþlzð Þ � a hklð Þ� �

where it is somewhat easier to recognize α(hkl) asan angle in the expression. The electron density ateach single point (x,y,z) in the unit cell is thus asummation of contributions from all the diffractionorders (h,k,l) in the diffraction photograph.The American physicist R. J. Havighurst had, in

1927, used this expression to calculate the electrondensity distribution in a crystal of sodium chloridealong cubic edges and cube diagonals.4 Bragg ex-tended the use of this method to two dimensions forthe structure determination of diopside and obtaineda clear solution for the position of oxygen atoms.5

Now one absolutely necessary piece was built intothe ground for determining the structures of largermolecules, especially of molecules as large as pro-teins. However, two other important pieces re-mained to be found: the diffraction data from pro-tein crystals and the phase angles, |F(hkl)|obs andα(hkl), respectively, in the above expression for theelectron density.

The Continued Road towards ProteinStructures

W. L. Bragg had begun his very successful X-rayanalysis of crystal structures in Cambridge. There-fore, it was logical that a post as lecturer in structuralcrystallography was created there, in the Depart-ment of Mineralogy, in 1926. The person who wasappointed to this position was a young and talentedscientist, J. D. Bernal. After studies in mathematics,mineralogy, and geology in Cambridge, he beganresearch in 1923 in the laboratory of W. H. Bragg


(the father), who now had moved to the RoyalInstitution in London. Bernal became fascinatedwith crystallography, but his great interest was touse these X-ray techniques on biologically interest-ing molecules. Thus, in the Department of Miner-alogy, he studied some minerals, but his maininterest was to investigate molecules of biologicalimportance, even some proteins. How reasonablewas that, when so far it had not even been possibleto obtain any diffraction data from protein crystals?However, Bernal was a visionary and optimist.

Furthermore, laboratory facilities were mostlikely improved when crystallography was trans-ferred to the famous Cavendish Laboratory in1931. Then in 1934, Bernal's group happened toobtain some very nice crystals of the protein pepsin.They were brought from Prof. Svedberg's labora-tory at Uppsala University, Sweden (my homeuniversity). Svedberg had developed the ultra-centrifuge, which was an excellent instrument forseparation of macromolecules. A visiting scientistfrom Oxford came to Uppsala with a sample con-taining pepsin molecules in order to prepare a puresample of the protein. When the separation hadbeen performed, he left the pure pepsin solutionin a test tube, where very nice crystals grew aftersome time.A friend of Bernal who shortly afterwards visited

Svedberg's laboratory and saw the crystals was surethat Bernal would love to look at them, and he gotsome of them to bring to Cambridge. When Bernalexposed these crystals to X-rays, they gave no re-flections, only a vague blur—the same negativeresult that other people had obtained using proteincrystals. However, Bernal did not give up. He exam-ined the crystals under a microscope and discoveredthat when they were removed from their motherliquor, they deteriorated, became white, and lostbirefringence. Bernal realized that the crystals had tobe kept in their mother liquor; therefore, hemounteda crystal wet in a thin-walled glass capillary. Itremained clear and, when exposed to X-rays,produced an excellent diffraction pattern; the filmwas covered with a mass of reflections. A youngvisiting scientist from Oxford, Dorothy Crowfoot(later Hodgkin), helped Bernal with the character-ization of X-ray films by determining unit celldimensions, space groups, etc. In May 1934, theypublished in Nature the first X-ray photograph froma protein crystal.6 When this work was finished,Dorothy Crowfoot returned to Oxford, where shemade an outstanding contribution to structurebiology by solving the X-ray crystal structures ofcholesterol, penicillin, vitamin B12, and, finally,insulin. Thus, together with her colleagues in Cam-bridge (see Chapters 2 and 3), she became one of thefounders of protein crystallography.Now only one vital piece of information—the

phase angles—remained missing in the process ofmaking protein structure determinations possible.Finding a solution to the phase-angle problem inprotein crystallography required almost another20 years.

Perutz and Kendrew, and Their Work toReach the Goal

In 1936, Max Perutz, a young research studentfrom Austria, joined Bernal's group. Perutz wasborn in Vienna in 1914 and studied chemistry at auniversity there. During a lecture in organicchemistry, Max heard about work on vitamins andenzymes going on in Cambridge (in the Departmentof Biochemistry), and he became certain that Cam-bridge was where he wanted to work for his docto-rate. As it so happened, Max never started hisresearch studies in biochemistry, but instead inBernal's group at the Cavendish Laboratory. There,biologically important molecules were studied,attracting Max even though the method was crystal-lography about which Max knew nothing.When Max started to work in Bernal's group,

there was no useful biological specimen availablein the laboratory. Therefore, he had to begin withwork on some minerals, which was somewhatdisappointing. However, Max was still happy towork in the laboratory with his visionary leaderwho always was optimistic about the power ofX-ray methods. The biological specimen that Max somuchwanted to work on was finally obtained by hisown action. A relative who was a physical chemistrecommended that he should begin structuralstudies on haemoglobin and also suggested some-body in Cambridge who might be able to supplyhim with crystals of haemoglobin. This worked outwell, and Max got beautiful crystals of the proteinthat colleagues in the group helped him to mountand start X-ray studies on. Max determined thecrystals' basic parameters, unit cell dimensions, andspace groups. Furthermore, the high quality of theX-ray pattern clearly indicated that the structure ofthe protein molecules in the crystals was welldefined, but how to proceed further was, of course,not obvious.In 1937, Bernal accepted a position as professor at

Birkbeck College in London, and the members ofhis group were invited to go with him there.However, Max decided to stay since he liked bothCambridge and the Cavendish Laboratory verymuch. At the same time, the old Cavendish Pro-fessor, Rutherford, died, and his successor hap-pened to be W. Lawrence Bragg. Max had, ofcourse, learned that Bragg was the inventor ofX-ray crystallography, and he was both nervousand hopeful when he called on Bragg to describewhat he worked on and to show the X-ray photo-graphs from the haemoglobin crystals. Bragg be-came very glad and enthusiastic. He immediatelysaw the great possibility that the method he hadinvented could one day be used to solve impor-tant problems in biology and medicine. Shortlyafter this meeting, Bragg managed to obtain agrant from the Rockefeller Foundation, and Maxwas appointed as his research assistant. Max hadreally found the ideal supporter and mentor.Bragg was highly respected and knew extremely


well both the powers and the difficulties of X-rayanalysis.The work on haemoglobin was interrupted by the

war. Max was first interned in Canada as an “enemyalien” (from Austria) and then was invited to join asecret war project that aimed at producing a floatingairbase from ice, a work that never came to anypractical application. (Max, an avid skier, had pub-lished scientific papers on the structures of ice.) Atthe end of the war, Max returned to Cambridge and,in 1946, was joined by John Kendrew.Kendrew was born in Oxford in 1917. After basic

studies there, he went on to Trinity College in Cam-bridge, where he graduated in Chemistry in 1939. Hethen started research in physical chemistry (reactionkinetics) but, after a few months, was called up formilitary service, where he, among other things,worked on radar. During military service in South-east Asia, Kendrew happened to meet J. D. Bernal,who explained to him the great and challengingpossibility of determining protein structures bymeans of X-ray analysis. Bernal's description of hisresearch no doubt had a big influence on Kendrew,and this interest in protein crystallography was fur-ther amplified during a visit to California, where Johnmet Linus Pauling. It was natural that Johnwanted tofinish his Trinity scholarship in Cambridge, where hewas accepted to work on haemoglobin together withMax. It was decided that John should not work onexactly the same form of haemoglobin as Max (theprotein from horse), but instead compare the proteinforms from adult and foetal sheep.Thus, there was a good plan for the scientific

work. However, the financial situation was moredifficult. Both John and Max had grants and awardslasting only another 2 years. Bragg had tried to get auniversity lectureship for Max, but the decision onthis request seemed to last extremely long. Max'sremark on this was that he probably was considereda very odd scientist—a chemist in a Physics Depart-ment working on a biological problem. Today, wewould, of course, say that this is an ideal situationfor somebody working in “molecular biology.”

The Medical Research Council Unit

The financial difficulties of Max and John weresolved in an interesting way. John andMax had beengiven bench space in a parasitology institute inCambridge, the Molteno Institute, headed by DavidKeilin. Max told Keilin about the difficult finan-cial situation, and this opened the way towards asolution. Keilin was a good friend of Sir EdwardMellanby, the Executive Head of the Medical Re-search Council (MRC), and Keilin suggested thatBragg should discuss with Mellanby the possibilityof getting governmental financial support from theMRC. After a meeting with Bragg, Mellanby pre-sented a written suggestion to the Council and, intheir meeting in October 1947, the establishment ofthe “Medical Research Council Unit for the Study ofthe Molecular Structure of Biological Systems” in

Cavendish Laboratory was decided. The grant was£2550 per year to support Max and John and tworesearch assistants for 5 years. This was the start of aresearch laboratory that would grow into a veryfamous research institute. The interesting thingabout this important decision was that the physicistBragg managed to convince the MRC that thisresearch could lead to important results of greatbiological and medical value.

The Final Solution: The Phase-AngleProblem Solved

After Kendrew had finished his doctorate, hedecided to change project from haemoglobin to thefour-times-smaller protein myoglobin. The twomolecules were closely related. Haemoglobin,with its four protein chains and haem groups, hasthe role of transporting oxygen to cells, while thesingle-chain myoglobin, with only one haem group,stored oxygen in muscles (serving as a reserve foroxygen). Kendrew chose this new project mostprobably because it would be easier to determinethe structure of this smaller molecule and, with theirfunctions being so closely related, there naturallywere good possibilities for interesting comparisonsof the two structures.It was, of course, a great help that the financial

situation was secure, especially since the initial yearsturned out to be scientifically difficult. Of the manymethods used in trying to get structural results,probably the best was to vary the salt concentrationin haemoglobin crystals. This led to some informa-tion on phase angles in case these were 0° or 180°.That made it possible to calculate a projection of thehaemoglobin molecule, which, however, wasimpossible to interpret.Then in 1953, Max started to work along a line

that had been discussed and even used many yearsearlier. The method was to incorporate a heavyatom into the molecules, crystallize the modifiedmolecules, and thus obtain somewhat changed X-ray reflection intensities. This method had beenused successfully on smaller molecules (e.g., by M.Robertson in Glasgow, where Michael Rossmann,the author of Chapter 3 of this paper, did his PhDwork). However, nobody thought that it should atall be possible to apply this method to such hugemolecules as proteins. Even a heavy atom with, forexample, 80 electrons was not expected to producea measurable change in the diffraction pattern froma molecule containing a thousand or more lighteratoms. However, Max did not give up. Instead, hemade a very important experiment. He measuredon haemoglobin crystals the so-called absoluteintensities (i.e., the intensities of diffracted X-raysrelative to the intensities of the incident beam). Theresult was that these absolute intensities weresurprisingly weak. The scatterings from the manylight atoms apparently tended to cancel each otherto a great extent (see Fig. 2.1c in Chapter 2 of thiswork).

Fig. 1.3. John Kendrew photographed ca 1962, the yearhe shared the Nobel Prize in Chemistry with Max Perutz.


Furthermore, Max obtained interesting informa-tion from a biochemist at Harvard who investigatednormal and sickle cell haemoglobin to look fordifferences. During these studies, he could see that itwas possible to attach mercury atoms to sulphydrylgroups on haemoglobin without any change in theoxygen uptake. This indicated that the structure ofthe molecules most likely had not been altered, agood sign for Max. Vernon Ingram, a good chemistin the group, now prepared a mercury compoundthat he attached to the sulphydryl groups of haemo-globin. Max succeeded in crystallizing this haemo-globin “derivative” and exposed one of thesecrystals to X-rays. When he developed the photo-graph, he could, even when the film was still wet,see clear changes in reflection intensities comparedto the intensities from crystals of the unmodifiedprotein. He immediately rushed to Bragg's officeand brought him to the dark room. They looked atthe two films and at each other, and, almost withoutsaying anything, they both understood that this wasthe solution.Fifteen years earlier, they had together looked at

one film. Now they had two. The present “deriva-tive” film itself was not enough. Mercury atoms didnot dominate but merely altered the intensities onthe “unmodified” film, producing small but mea-surable differences that could be used for the deter-mination of phase angles. In fact, two different deri-vatives were needed (even theoretically) to calculateunique phase angles. These derivatives were quitepossible to prepare even if some effort was needed.

Fig. 1.2. Max Perutz photographed by Dan Andersonin 1991 during a visit to the University of California atLos Angeles.

The way to determine the phase angles by means oftwo or more derivatives is well explained in the textand in Figs. 2.2 and 2.3 (Chapter 2) of this work.Since the crystals of the derivatives and the unmo-dified protein must be isomorphous with one ano-ther, the method was called isomorphous replacement.Max and his colleagues published this important

result in 1954 in the Proceedings of the Royal Society,7

and now Max (Fig. 1.2) and John (Fig. 1.3) began to“see the light at the end of the tunnel.”

The First Picture of a Protein Molecule:Myoglobin at 6 Å Resolution

The way towards a first low-resolution structurewas not surprisingly easier and faster for the smallermyoglobin. The type ofmyoglobin specimen used byKendrew, from sperm whale, could be obtained inlarge quantities and gave large and robust crystals. Alow-resolution analysis was tried first, since thisinvolved a reasonable amount of data to be handledin this first step. The resolution of 6 Å was chosenbecause, at this level, it should be possible to recog-nize polypeptide chains having a compact config-uration such as a helix. (In 1951, Pauling et al.8 hadpredicted the existence of α-helix configuration inprotein molecules, and Max had later that yearproduced experimental support for this.)Preparation of heavy-atom derivatives for myo-

globin, which had no free sulphydryl groups, was


somewhat more difficult than that for haemoglobin.The two chemists Gerhard Bodo and HowardDintzis in John's group prepared and tested a largenumber of derivatives and succeeded in finallyobtaining five compounds whose crystals gavereasonable intensity changes compared to crystalsfrom the unmodified protein. The positions of theheavy atoms were determined by using a functionderived by Patterson.9 A “Pattersonmap,” for whichknowledge of phase angles was not needed, gaveinformation on the relative positions of the heavyatoms in the different derivatives, but not on theatom positions themselves. For myoglobin, anacceptable but nonoptimal positioning of the heavyatomswas obtained in this way. (During thework onhaemoglobin to 5.5 Å resolution, Michael Rossmann,the author of Chapter 3 of this work, derived aperfect solution to this problem.) Now all informa-tion for determination of the phase angles wasavailable, and this was performed graphically withhand-drawn phase-circle diagrams as shown andexplained in Fig. 2.4 of Chapter 2. Finally, theelectron density was calculated using the 400 mea-sured intensity data from unmodified crystals andthe corresponding phase angles. These calculations,which took a total of 70 min, were performed on theCambridge University EDSAC I computer.From the calculated electron density maps, a

model of the myoglobin molecule to 6 Å reso-lution was constructed (Fig. 1.4). The model con-tained a number of dense rod-like features thathad the dimensions of α-helices and made up thebulk of the polypeptide chain. The rods were joinedby corners where the electron density, in most cases,

Fig. 1.4. The molecular model of myoglobin at 6 Åresolution. The model contains a number of dense rod-likefeatures that have the dimensions of α-helices andmake upthe bulk of the polypeptide chain. The chain is folded in anirregular manner, more complicated than had beenanticipated. In a pocket in the molecule is a dense flatteneddisc, presumably the position of the hemegroup. Reprintedwith permission from The Royal Society, London.

was somewhat lower or even broken. The chain wasfolded in an irregular manner, certainly much morecomplicated than had been anticipated. In a pocketin the molecule (having approximate dimensions of45 Å×35 Å×25 Å), there was a dense flattened disc,which presumably was the position of the hemegroup with its central iron atom.This result was ready by the beginning of 1957

and was published in Nature in the following year.10

It was the very first picture of a protein molecule,and it was followed by the wonderful structuredeterminations of myoglobin to 2.0 Å and of haemo-globin to 5.5 Å, performed in 1959 and described inChapters 2 and 3, respectively.

Working on a Protein Molecule inCambridge: A Privilege and a GreatExperience

I started research in the Department of Generaland Inorganic Chemistry at Uppsala University in1955. Since the professor at the department, GunnarHägg, was among the three or four people whointroduced X-ray crystallography in Sweden, thiswas the key method at the institute. My supervisorIngvar Lindqvist, who was interested in metalcomplexes, gave me the compound AgNH4(SCN)2as my first structural problem. I thought this com-pound was not so exciting, but a nice experienceduring this work was to use a machine for thecalculation of Fourier series, recently constructed byGunnar Hägg himself. Our building was part of abig complex that included all branches of chemistry.When we went for coffee in the morning, we passedby the Physical Chemistry Department (from whereBernal's pepsin crystals, produced with the help ofSvedberg's ultracentrifuge, came in 1934) andalmost reached the Biochemistry Department,where Arne Tiselius, the man who developed theelectrophoresis method, was the professor. Coffeetimes were therefore excellent occasions to learnabout problems in other branches of chemistry andresulted in my second and third structural problemsbeing related to biochemistry. These were the metalcomplexes Cu(glycylglycine)2 and Zn(imidazole)2,with the latter being a model compound for theactive center of the enzyme carbonic anhydrase,which was studied in the Biochemistry Department.This was something quite different and, duringwork on these structures, I also attended a course inbiochemistry in Tiselius' department.Thus, it was very exciting when Gunnar Hägg

in May 1957 received a letter from Max Perutz,asking if somebody in Hägg's department might beinterested in coming to Cambridge and taking partin continued work on myoglobin or haemoglobin. Iwas, of course, immediately very interested, butthere was a small problem. I had not yet finished myPhD work. On the other hand, such an opportunitywould probably not come again, so I was given theright to make a break in my PhD studies. Hägg


wrote back to Max and explained my backgroundand my big interest in joining Max and John. Thus, itwas decided that I should go to Cambridge, butsince our first child was recently born, it was first inJune 1958 that I traveled to England together withmy wife Karin and my little son Peter.When I arrived, Max asked me which of the two

problems I wanted to work on. Both were, of course,very interesting. Taking part in work on the firsthigh-resolution structure of a protein moleculeseemed an irresistible challenge, so I started towork in John's group and became a close colleagueof Dick Dickerson, who already had been workingfor some months on the 2Å structure of myoglobin.The laboratory, “The Hut,” was much smaller thanthe Department of Chemistry from which I hadcome, but the atmosphere was wonderful. I came toshare a small office with Dick; Michael Rossmann,who arrived a few months later and started to workon haemoglobin; and Larry Steinrauf, who was apostdoctoral fellow from Linus Pauling's laboratory.It was really a great privilege to share an office withthese wonderful colleagues. All the people in thelaboratory—such as Mary Pinkerton, who was thehead of the staff measuring the large amount ofX-ray data; Ann Cullis, who was assistant to Max;

Fig. 1.5. Group picture of some of the people working in ThSteinrauf, Richard Dickerson, Hillary Muirhead, Michael RoStrandberg, Wibeke, and an unidentified person. Front row: L

Hilary Muirhead, who was a research student in thehaemoglobin group; and many others—were verykind and helpful. The leaders Max and John wereexcellent examples for us, as were Francis Crick andSidney Brenner, who shared an office across thecorridor from our office. Francis was the great leaderduring the 11 o'clock coffee break, when all possiblesubjects were discussed. Figure 1.5 is a grouppicture from the autumn of 1958, with some of thepeople working in The Hut gathered outside theentrance door of the small laboratory building.To me, three things were special with this labo-

ratory and with the people working there:

(1) Everybody was always keen to solve all up-coming problems at any stage of the work,and this was, of course, very natural withleading scientists such as Max and John.

(2) The equipment was top class for that time.X-ray generators with rotating anodes (con-structed by the laboratory engineer TonyBroad); a microdensitometer for measuringthe vast number of X-ray reflections; and theEDSAC II computer at the nearby Mathe-matics Laboratory (probably the fastest com-puter in the world then) were all facilities

e Hut (autumn of 1958). From left to right, back row: Larryssmann, Philip ___?___ (face obscured), Ann Cullis, Broreslie Barnett, Mary Pinkerton, and Max Perutz.

†Ray Pepinsky, by all accounts, was a tyrant in thelaboratory and a rugged taskmaster. A postdoctoralfellow from the Netherlands whom I knew in Lipscomb'slaboratory had moved from Penn State to Minneapolis


that strongly contributed to making the workpossible.

(3) Finally, the many distinguished visitors cer-tainly stimulated us and made us feel that wewere working on very important problems.

My biggest memory from this time was no doubt asunny morning in the beginning of August 1959,after a whole night of calculating the electron den-sity map of myoglobin. In Chapter 2, Dick has givena very nice description of that event.After taking part in the model building of myo-

globin, I went back to Uppsala in December 1959and started about a year later to work on the three-dimensional structure of the enzyme carbonic anhy-drase, which was solved in 1969.

References

1. Friedrich, W., Knipping, P. & Laue, M. (1912).Interferenz-Erscheinungen bei Röntgenstrahlen. Sit-zungsber. Kgl. Bayrischen Akad. Wiss. 303–322.

2. Bragg, W. L. (1913). The structure of some crystals asindicated by their diffraction of X-rays. Proc. R. Soc.London Ser. A, 89, 248–277.

3. Bragg, W. L. & Warren, B. (1928). The structure ofdiopside, CaMg(SiO3)2. Z. Kristallogr. 69, 168–193.

4. Havighurst, R. J. (1927). Electron distribution in theatoms of crystals. Sodium chloride, and lithium,sodium and calcium fluoride. Phys. Rev. 29, 1.

5. Bragg, W. L. (1929). The determinations of parametersin crystal structures by means of Fourier series. Proc.R. Soc. London Ser. A, 123, 537–559.

6. Bernal, J. D. & Crowfoot, D. (1934). X-ray photographsof crystalline pepsin. Nature, 133, 794–795.

7. Green, D. W., Ingram, V. M. & Perutz, M. F. (1954).Structure of haemoglobin: IV. Sign determination bythe isomorphous replacement method. Proc. R. Soc.London Ser. A, 225, 287–307.

8. Pauling, L., Corey, R. B. & Branson, H. R. (1951). Thestructure of proteins: two hydrogen-bonded helicalconfigurations of the polypeptide chain. Proc. NatlAcad. Sci. USA, 37, 205–211.

9. Patterson, A. L. (1935). A direct method for the deter-mination of the components of interatomic distancesin crystals. Z. Kristallogr. 90, 517–542.

10. Kendrew, J. C., Bodo, G., Dintzis, H. M., Parrish, R.G., Wyckoff, H. & Phillips, D. C. (1958). A three-dimensional model of the myoglobin molecule ob-tained by X-ray analysis. Nature (London), 181, 662–666.

after a nasty row in which Pepinsky had threatened to

Chapter 2: Myoglobin: A Whale of aStructure!

Richard E. Dickerson

have his student visa revoked and to send him back homeagain. A few years later, as a new faculty member atCaltech, I mentioned Ray Pepinsky to a faculty colleague,Eddie Hughes. Hughes was the most courteous, gentle-manly, and mild-spoken academic whom I can recall. Inever heard Eddie say a bad or spiteful word aboutanyone, no matter what the provocation. I told Eddie that
I had heard that Pepinsky had just moved from PennState to Nova University in Florida, and that Penn Statehad allowed him to take with him the vast X-RAC facilitythat he had built up. When I moved from the Universityof Illinois to Caltech in 1963, I brought almost noequipment with me. “How did Pepinsky manage to getPenn State's permission to walk off with many thousandsof dollars of computer equipment?”, I asked. Eddiesurprised me with a calm reply, “Well, I suppose theyfigured that was the price they had to pay.”
Introduction

How is a crystal put together? The picture in manypeople's minds is of atoms or ions packed tightlytogether like tennis balls in a large cardboard box.Some crystals have only one type of ball, such ascarbon atoms in diamond. Other crystals have two

different kinds, such as sodium and chlorine ions inNaCl. But crystalline proteins or crystalline DNA?Ridiculous!In 1953, I joined the laboratory of Bill Lipscomb at

the University of Minnesota to work on crystalstructures as a step toward a doctorate in physicalchemistry. When I arrived, on the “Colonel's”(Lipscomb was from Kentucky) desk was a weirdskeleton of linked triangles forming a fragment of anicosahedron, with one half open like a footballhelmet. This was the structure of a boron hydride,the Colonel told me, with boron atoms at the cornersof the triangles and hydrogen atoms attached inunexpected places. I remember being astoundedthat such an awkward and ungainly entity couldever be induced to form a crystal. Little did I realizethat in Cambridge, England, two colleagues namedPerutz and Kendrew were, at that very moment,well on their way toward solving the crystalstructures of molecules of vastly greater complexitythan mere boron hydrides.Things were much simpler in those distant days. I

earned my doctorate with the aid of 375 X-rayreflections by solving the crystal structure ofB9H15.

1–3 The project had been begun by a visitingprofessor from Leeds University, Peter Wheatley,and I inherited it when he returned to England. Datawere collected on film with a precession camera,spot intensity was gauged by visual comparisonwith a standard scale, and calculations were carriedout on a desk calculator with the help of an IBM 407punched-cardmachine, which could perform simplearithmetic operations on data entered via punchedcards. IBM 407 was “programmed” by connectingwires on a plugboard, rather like a primitive tele-phone switchboard. For the production of electrondensity maps, punched-card data were sent to RayPepinsky at Penn State University, who had devel-oped the X-Ray Analogue Computer or X-RACspecifically for crystal structure analysis†.3


Things improved vastly when the Colonel per-suaded Honeywell Corporation in nearby St. Paul tolet us use their Remington Rand UNIVAC 1103Acomputer after office hours. The University ofMinnesota had no digital computers of any kind atthat time. They had considered purchasing an IBM650 magnetic drum computer but decided that therewas too little demand for its services to justifyspending US$200,000–400,000 on such a machine.The UNIVAC 1103A was a “huge” computer, with1024 thirty-six-bit words of electronic memory and16,384 words on a magnetic drum. Compare thatto 2 GB or 2,000,000,000 bytes of storage on aremovable flash drive memory stick for a modernlaptop computer. One byte in modern usage corres-ponds roughly to 8 bits, so today's 2-GB laptopmemory stick has more storage capacity than 25,000UNIVAC 1103A computers! As I was finishing mydegree, a young postdoctoral fellow from Glasgow,Michael Rossmann, showed up on the scene. He andAudrey, and my wife Lola and I became goodfriends, but when I left Minneapolis in February1957, I never dreamed that, in only a couple of years,we would be working together again, a third of theway around the world.I did not go straight to Cambridge. In fact, the

Colonel told me that Oxford and Cambridge wereold hat, and that I would be better off at LeedsUniversity, which offered such ground-breakingcrystallographers as Peter Wheatley, DurwardCruickshank, and George Jeffrey. So one cold dayin February of 1957, Lola and I disembarked atSouthampton and took a train north to Leeds, wherePeterWheatley met us at the station. But things werenot what they might have been. Jeffrey had finallymoved permanently to the University of Pittsburghwhere, over the years, he built up a large andproductive Graduate Crystallography Department,the only one of its kind in theUnited States. DurwardCruickshank hadmoved toManchester, which had adigital computer (the Ferranti Mark I) that LeedsUniversity lacked. (When Leeds obtained its firstcomputer a few years later, the university housed itin a former Methodist church building adjacent tothe campus. In a bizarre fit of inspiration, its officialname was the “Leeds University Computer andIntegrator, Ferranti” or LUCIFER.)At Leeds, I collected precession data on a now

mercifully forgotten dimethylsulfoximine, had aPatterson function calculated on the Manchestercomputer, and then solved the three projections ofthe structure using Beevers–Lipson strips and adesk calculator. But during the summer of 1957, acrisis arose. As I have related before,4,5 Wheatleyannounced that he could no longer support a wifeand two children on his annual academic salary of£900 and was leaving Leeds to head up a newMonsanto crystallographic laboratory in Zurich.By a stroke of extraordinarily good fortune, at that

very moment, Max Perutz and John Kendrew inCambridge were advertising for one or more post-doctoral fellows to work with them on the high-resolution crystal structure of myoglobin and the

low-resolution structure of the larger haemoglobin.Perhaps in a fit of conscience, Bill Lipscomb pro-posed my name and provided a very positive re-commendation. Peter Wheatley also suggested me,and the result was that I received a phone call fromMax, inviting me down to spend a day talking andlooking around the laboratory. During the day, weall visited the Royal Institution in London to talkwith David Phillips, who was collaborating with theCambridge team. It was there that I first met awonderful person, Sir Lawrence Bragg, the previousdirector of the Cavendish Laboratory at Cambridgeand now the director of the Royal Institution. It wasBragg who, together with Max Perutz, had built theMedical Research Council (MRC) Laboratory inCambridge and attracted to it people such as JohnKendrew, Francis Crick, and Jim Watson. In 1915,the 25-year-old Bragg had shared the Nobel Prize inPhysics with his father, W. H. Bragg, for essentiallyinventing X-ray crystal structure analysis. I lookedat the man who had created Bragg's Law and sup-pressed an urge to exclaim, “I assumed you weredead!”There was a meeting of the minds, and I accepted

their invitation to join the MRC group at Cambridgein the fall of 1957. I chose to work with JohnKendrew in part because the idea of getting a struc-ture at atomic or near-atomic resolution appealed toa small-molecule crystallographer.

Getting Started

The MRC Laboratory of Molecular Biology occu-pied a long, one-story temporary building that hadbeen built as a metallurgical laboratory duringWorld War II. It was known irreverently as “TheHut” and sat in a courtyard of the CavendishPhysics Laboratory. The Hut contained a wetlaboratory and offices, but the X-ray machinesthemselves were mercifully housed in the basementof the adjacent and more sturdily constructedCavendish Laboratory. The wet laboratory was atone end of The Hut. One side of the central corridorhad offices for Max Perutz and for Francis Crick andSidney Brenner. On the other side were the officesfor John Kendrew and for postdoctoral fellows.When I arrived in the fall of 1957, I shared this officewith Erwin Alver and Roger Hart. Alver, from theUniversity of Oslo, was an inorganic crystallogra-pher like myself. Roger Hart was somewhat morequalified: although originally from the UnitedStates, he had first worked with Rosalind Franklin,and then with Aaron Klug after her death.Alver only stayed for 1 year and returned to Oslo

in the summer of 1958, shortly after the arrival of athird colleague, Bror Strandberg, from Uppsala.Alver worked with Kendrew like myself, and Hartworked with Perutz. Prior to Bror's arrival, Alverand I joked that we were the only two people in theMRC Laboratory who had ever completely solved acrystal structure, although mine had only nineboron atoms and Alver's was comparably insigni-


ficant. Proteins constituted an entirely new universefor both of us.Kendrew had initially studied horse myoglobin,

but soon reasoned correctly that the oxygen-storingmolecule in tissues should be especially plentiful inaquatic mammals that dived and remained underwater for appreciable time periods. Finback-whaleand sperm-whale myoglobins were isolated andtried, and the latter was selected because itsmyoglobin concentration was so high that themeat was almost black.6 Sperm-whale meat wasshipped from Peru and stored frozen in large gar-bage cans in a deep-freeze room of the MoltenoInstitute, down the road a piece from the MRCLaboratory. Myoglobin was isolated and purified bylaboratory aides, and crystals were grown in 2- or4-ml screw-top vials. Crystals grew so large that theyhad to be cut down to size with a scalpel blade beforebeing inserted and sealed in glass capillaries for X-rayphotography.Roger Hart was unexpectedly an important source

of inspiration to me in these early days. John Ken-drew carefully demonstrated to me every step of thecrystal-mounting process: choosing a good crystal,cutting it down to size, inserting it into the capillaryand drying away excess solvent from around it,adding short plugs of solvent to both ends of thecapillary to maintain a moist atmosphere, sealingboth ends of the capillary with wax, and mounting iton a goniometer head for placement on a precessioncamera. I found this procedure incredibly difficultand became rather intimidated by the entire process.I began to wonder whether I would ever master thetrick. But Roger gave me pause. His hands weretwisted and crippled by polio, yet he was takingsurvey picture after survey picture of derivatives ofhaemoglobin for Max Perutz. I said to myself, “ByGod, if Roger can mount crystals with those hands, Imost certainly should be able to.” And so it was,after hours of trial and error. Only later did I learnthat Roger did not actually mount the crystals thathe was photographing. Perutz came in after hoursand mounted the next day's supply of haemoglobincrystals!Bror Strandberg arrived from Uppsala in June

1958, and the project really got going. The thirdmember of our myoglobin team was David Phillipsin London. He was to collect data for one of theheavy-atom derivatives at the Royal Institution.John's original plan had been to follow his 6-Å low-resolution analysis with a higher-resolution analysisat 2.5 Å. The 6-Å analysis required ca 400 indepen-dent X-ray reflections from parent myoglobincrystals and from each of the five heavy-atom deri-vatives. Increasing the resolution from 6 Å to 2.5 Åwould necessitate collection of (6/2.5)3 ×400=5500reflections from each crystal type. But in planningdiscussions, someone pointed out that, at 2.5 Å, wewere not using the full power of the precessioncamera. The greatest resolution that we could obtainon our 5 in.×5 in. photographic plates was 2.0 Å, forwhich the reflections would extend all the way outto the edges of the plate. Why not go for broke and

collect 2.0-Å data instead? We decided to do so andto collect and measure ca (6/2)3 ×400=10,800 reflec-tions per data set. This would give us twice theinformation as the originally planned 2.5-Å analysis.(In the end, after allowing for space group extinctreflections and other problems, a dependable set of9600 reflections was used in the 2-Å analysis.)Data were collected from four heavy-atom deri-

vatives: para-chloromercuribenzene sulphonate,mercury diamine (HgAm2), gold chloride, and thedouble derivative para-chloromercuribenzene sul-phonate/mercury. Each data set required 22 differ-ent precession camera photographs; for eachphotograph, a newly mounted myoglobin crystalwas used to minimize radiation damage. In pre-cession cameras, X-ray reflections or spots arearranged in a series of parallel lines. This meantthat one could use a Joyce–Loebl microdensitomer toscan along each of these lines, yielding peaks thatthen could be measured with a millimeter scale. Maxand John had a crew of technicians who were avail-able to carry out this process, and we ended with along list of (h,k,l) reflection indices and their corres-ponding peak intensities. These then were typed onpunched paper tape for input to the Electronic DelayStorage Automatic Calculator (EDSAC) II computerin the nearby Cambridge Mathematical Laboratory.EDSAC II was considered a major step up from

the EDSAC I that Kendrew had used for low-resolution myoglobin. Kendrew and his colleagueBennett in 1952 were actually the first people toprogram a digital computer or “digital electroniccalculating machine,” the EDSAC I, to carry out athree-dimensional Fourier synthesis.7 Bennett sub-sequently moved to the Ferranti Corporation inManchester, the place that also lured DurwardCruikshank away from Leeds. This new EDSAC IIcomputer had 2000 words of high-speed corestorage and two magnetic tape drives. The corewas far too small to permit shuffling and reorderingof 10,000 X-ray reflections from each of five sets ofdata: a parent protein and four derivatives. Mergingof data had to be accomplished by thumb-tackingstrips of punched paper tape containing intensitydata to a bulletin board, shuffling the stripsmanually to merge the data sets, and then runningeverything back into the computer and emitting onegrand master tape. Figure 3.7 of Michael Ross-mann's chapter shows Bror and me walking ourdata bulletin board back from the computing center.A good strong wind might have made a majorchange in our analysis.Input to EDSAC II was performed by punched

paper tape using standard British telegraph readers.Little metal fingers in the reader pushed up throughthe holes in the tape, reading the information in thepattern of holes. If an input tape was mispunchedduring preparation, it was ruined and could only becorrected by making a new tape. In desperation, thecustom of patching the occasional mispunched holewith Scotch tape was developed. This worked fine,unless the tape reader fingers pushed off the patch,causing a misreading and sometimes a tape jam. The

Fig. 2.2. For one X-ray reflection, addition of scatteringvectors from a parent molecule Fo and an added heavyatom Fx to yield the observed experimental scatteringfrom the heavy-atom derivative Fo+x.

9 Reprinted withpermission from the International Union of Crystallogra-phy (http://journals.iucr.org).

Fig. 2.1. For a given X-rayreflection, addition of waves fromdifferent atoms in one molecule. (a)A single atom X. (b) A three-atomX3 molecule, with similar atoms indifferent locations. The heavy arrowrepresents scattering for that reflec-tion from the entire molecule. (c)Result of tagging the three-atom X3molecule with a bromine atom. Be-cause the heavy atom has so manyelectrons, it dominates scattering,and its phase is very similar to thatfor the entire X3Br molecule.5


computing center came down on this with a hammer,threatening that any tape found to be patched in thismanner would be confiscated. (I never recall thisthreat actually being carried out.)The heart of EDSAC IIwas an array of long racks of

vacuum tubes, with each rack being capable of beingpulled out individually by its handle when one of itstubes blew. Wiring diagrams for the computer werein pencil in a large stack of circuit diagrams. When achangewasmade, the diagramswere erased and cor-rected. Whether permanent or even inked drawingsof the computer existed, I was never aware of them.There were three grades of access to EDSAC II:

users, partially authorized users, and fully author-ized users. Users could only have normal access tothe computer during the day while it was mannedby computer center staff. Partially authorized userswere allowed to run the computer as late as theychose and to turn it off when finished by pulling aset of wall switches in a defined order. Fully autho-rized users were permitted to turn the computer onand off. I eventually qualified as a partially autho-rized user. I do not know if any member of theprotein crystal structure group ever rose to the rankof a fully authorized user, but if so, it probably wasMichael Rossmann. His programs for differencePatterson and difference Fourier refinements wereso complex that they were sometimes used for the

morning computer check, instead of the standardsystem-checking routines.

I Don't Recall the Name, But the PhaseIs FamiliarWe have not yet considered why five different

sets of data, each of roughly 10,000 reflections, were

Fig. 2.3. Phase-circle diagram for one X-ray reflection,using experimental amplitudes from the parent moleculeO and the two heavy-atom derivatives X and M. Eachderivative phase circle crosses the parent circle O at twoplaces. The intersection that is common to the twoderivatives is the correct phase angle.9 Reprinted withpermission from the International Union of Crystallogra-phy (http://journals.iucr.org).


necessary. This brings us to the phase problem,which is the key to the structure analysis of macro-molecules. One can generate an electron densitymapof the protein by a process of Fourier synthesis. Eachof the thousands of reflections in the X-ray patterncontributes onewave to the image of the protein. Thedirection of each wave is determined by the positionof its reflection in the X-ray pattern, and the wave-length of the wave is inversely proportional to thedistance of the reflection from the center of the X-raypattern. The amplitude of each wave is the squareroot of the measured intensity of the X-ray reflection.

Fig. 2.4. Phase circles for three X-ray reflections from myodifferent heavy-atom derivatives. (a) An excellent, unambiguophase. (c) A poorly defined phase angle.11

So far, so good. This is nothing that a computercannot handle. But one more—and very serious—complication remains. Each one of these thousandsof waves has a phase—the fraction of a 360° cycle bywhich it is shifted compared to the other wavesbefore they are all added together. To calculate theimage of a protein, one must know both the ampli-tude and the phase of each of the 10,000 reflections.How does one find out the phase of each and everyreflection in the X-ray pattern?Each wave corresponding to one X-ray reflection

actually is the sum of waves from all of the atomsin the scattering molecule. This sum, again, is acomplex number. The amplitude of scattering fromeach atom depends on the number of electrons in theatom, and its phase depends on the position of theatom in the molecule. Figure 2.1a represents scat-tering from a single atom for one particular X-rayreflection. The length of the vector is the amplitudeof the wave, and the angle of the vector from thehorizontal or real axis is its phase. Figure 2.1b showshow scattering vectors from three atoms withsimilar numbers of electrons but different locationsadd together to produce an amplitude and a phasefor the three-atom molecule as a whole.One classic phase-solving approach for small

molecules is the heavy-atom method. Tag the mole-cule with a heavy atom having so many electronsthat it dominates scattering, as in Fig. 2.1c. First,locate the heavy atoms in the unit cell. (See MichaelRossmann's Chapter 3 for more on this issue.)Compute an electron density map using the phasesof the heavy atom itself, which you can calculate ifyou know where the heavy atoms are located in thecell. If you are lucky, this map will show the heavyatoms plus at least a partial image of the rest of themolecule. Add more of these atoms as they becomevisible in the maps, and iterate until you have thepicture of the complete molecule.

globin at 6 Å resolution, using parent myoglobin and fiveus phase determination. (b) A somewhat less well-defined

http://journals.iucr.org

http://journals.iucr.org

Fig. 2.5. Lack-of-closure errorsin phase analysis. In this diagram,the amplitudes of the protein andits two heavy-atom derivativesare represented by F, F1, and F2.Scattering from heavy atoms aloneis shown as vectors f1 and f2. Forderivative 1, the requirement thatF1=F+ f1 is satisfied at points A andB around the parent phase circle;for derivative 2, the requirementF2=F+ f2 is met at points B and C.Hence, the true phase angle isfound at the shared point B. Butfor the general phase angle ϕ, themeasured |F1| from derivative 1 istoo long by distance ɛ1, and that forderivative 2 is too short by distanceɛ2. These lack-of-closure errors canbe used to calculate the probabilityof correctness of any angle. At thecorrect point B, this probabilityrises, in principle, to unity.12 Re-printed with permission from theInternational Union of Crystallo-graphy (http://journals.iucr.org).

Fig. 2.6. Phase-angle probabili-ties plotted arounda 360° phase circleof unit radius. This particular exam-ple exhibits two Gaussian-like peaksof probability, one roughly twice theother. Vector m marks the center ofgravity of the probability curve.12

Reprinted with permission from theInternational Union of Crystallogra-phy (http://journals.iucr.org).



But it was not at all clear whether this approachwould work with proteins. Haemoglobin, with amolecular weight of 68,000, would have somethingon the order of 34,000 electrons around its atoms.How could one imagine that one or two mercuryatoms of 80 electrons each could dominate the

Fig. 2.7. Examples of phase determination for three X-ray re0.96 (a), 0.59 (b), and 0.05 (c), which was the most poorly determm in each diagram marks the most probable phase, and poprobability circle. In the single-peak function (a), the most prfunctions, the best phase is shifted away from m somewhatfrom the International Union of Crystallography (http://jour

phasing? Indeed, could one expect to see any changein X-ray intensities at all when 80 or 160 electrons areadded to a molecule that already contains 34,000? In1954, Max Perutz provided the Rosetta Stone forprotein crystal structure analysis when he demon-strated that, although heavy atoms did not dom-

flections in the 2-Å phase analysis. The figures of merit areined phase in the entire 2-Å myoglobin analysis. Position

sition b is the "best" phase using the centroid of the fullobable and best phases are the same. In the two bimodaltoward the secondary peak.12 Reprinted with permissionnals.iucr.org).

Fig. 2.8. Section through the electron density of thehaem group in the 2-Å structure analysis, with an idealframework superimposed. Positions of the “ears” extend-ing from the four five-membered rings are clearlyestablished.15


inate phases from a protein, they did at least alter theintensities of the reflections in a manner that mightallow one to deduce the phase angles of reflectionsfrom the parent protein.8 Classical heavy-atomphasing methods would not work, but isomorphousreplacement phase analysis might.The man who provided the methodology for this

kind of phase analysis was David Harker, who in1956 was trying to solve the structure of the proteinribonuclease at the Brooklyn Polytechnic Institute.9

Figure 2.2, reproduced from his work, shows howthe scattering vector from the protein plus heavyatom (Fo+x in his notation) is the vector sum ofscattering from the protein alone Fo and the heavyatom alone Fx. (Boldface represents vectors in phasespace, and regular font is used for vector ampli-tudes). The Harker phase-circle diagram in Fig. 2.3for two heavy-atom derivatives, X and M, is pro-duced by:

(1) Drawing a circle of radius Fo around the origin(2) Drawing a circle of radius Fo+x around a pointlocated at −Fx(3) Drawing a circle of radius Fo+M around a pointlocated at −FM.

Fo, Fo+x, and Fo+M are the measured amplitudesof X-ray scattering for that reflection from theparent protein and the two heavy-atom deriva-tives. Fx and FM are the scattering vectors forheavy atoms X and M alone, which you can calc-ulate, in both phase and amplitude, once you havelocated the positions of the heavy atoms in theunit cell. The location on the Fo circle where allthree circles intersect marks the correct phaseangle for the parent protein.Of course, nothing is ever as simple as it first

appears. Errors in intensity measurements produceerrors in the three phase-circle radii, which in turncan make phase-angle determination inexact. Thesolution is to collect data from three, four, or fiveheavy-atom derivatives instead of the minimal two,and to draw the best phase angle through theircloud of intersections. Because myoglobin is somuch smaller than haemoglobin, it was actually thefirst protein structure to be solved in three dimen-sions by Harker's isomorphous replacement me-thod.10,11 In the Kendrew Archives at the BodelianLibrary in Oxford, there resides a three-ring note-book in which John Kendrew drew a set of fourhundred 6-Å Harker phase-circle diagrams with aruler and a compass, using different-colored pencilsfor each derivative. When we began preparing for2-Å high-resolution phase analysis in 1958, Johnhandedme this notebook as a guide to what I shouldincorporate into a computer program. Today, thenotebook sits jealously guarded in the KendrewArchives at the Bodelian, and the photocopies that Iobtained recently are not of publication quality. Iwish that I had had access to a really good colorphotocopy machine half a century ago. But Figure2.4, from the published work, will have to do.

The most serious problem with phase analysis at2 Å was how to obtain information from poor phasedeterminations such as Fig. 2.4c. Several approacheswere tried.12,13 Michael Rossmann proposed andHilary Muirhead programmed a “least scatter” app-roach, choosing as a phase the radial line alongwhich intersections with phase circles lay mostclosely together. Roger Hart used a function to calc-ulate the probability of any particular phase angle,and then chose the direction that led to maximumprobability. But the optimal solution was developedby Blow and Crick.14 Using lack-of-closure errors asdefined in Fig. 2.5, they calculated the phase-angleprobability all around the 360° phase circle (Fig. 2.6),and then chose the centroid or center of gravity ofthat probability distribution as the “best” phaseangle. The vector to this centroid, m in Fig. 2.6, thenyields the best phase ϕB for that reflection in theFourier synthesis and the optimal weighting factorm. The length of that vectorm is 1 for a perfect phasedetermination and 0 for a hopelessly bad determina-tion. The length m itself is then a measure of thedependability of that particular phase angle and iscalled the “figure of merit.”12,13 The mean figure ofmerit provided a way of assessing the quality of theentire phase analysis.I wrote a program for EDSAC II that calculated

phase probabilities and defined the centroid vector.Figure 2.7 exhibits three examples from the high-resolution myoglobin analysis and shows theusefulness of the figure of merit m as a measure of


the quality of each phase determination. Calculationof phase angles was a slow process by modernstandards:

“With four heavy-atom compounds, the time re-quired for one phase determination on EDSAC IIwas 3.5 sec, of which about 1 sec was occupied inpunching results. The full set of 9600 reflectionsrequired 10 hr.”12

The mean figure of merit m ranged from 0.90 atthe center of the X-ray pattern (lowest resolution) to

0.45 at the high-resolution 2-Å limit. We experimen-ted with various exponential sharpening functions.One approach was simply to accept the figure-of-merit results, with its falloff of amplitudes in thehigh-resolution region of the diffraction pattern. Butthis damped down the calculated image too muchand smeared out fine structure features. The bestsolution was to apply a sharpening function so thatthe average falloff of the “best” data, includingfigures of merit, matched the average falloff of theoriginal experimental data set. This had the resultof keeping the effective resolution the same, but

Fig. 2.9. View through stackedPlexiglas sheets showing the haemgroup seen on edge and a viewdown the hollow F-helix that isattached to the haem by a His sidechain. Map, above; interpretation,below.15


emphasizing well-established phases and dampingdown those that were badly underdetermined.

Fig. 2.10. Density in an unrolled α-helix at 2 Å. Notonly are side chains clearly visible above and below theplane of this unrolled cylinder, the CfO groups along thebackbone are so well defined that they establish thedirection in which α-chain runs. The 6-Å “sausages” areclearly Pauling/Corey α-helices.15

The All-Night Party

For the last stages of myoglobin analysis—phaserefinement, electron density calculation, and inter-pretation of results—myoglobin had a three-manteam. David Davies from the National Institutes ofHealth inWashington, DC, joined us in April or May1959 and, like Bror, remained for a time after Ideparted for the University of Illinois in September1959. Once the weighted phase angles had beencalculated, the next step was to calculate the 2-Åelectron density map of myoglobin and see whatnew information could be learned. The main ques-tion was whether those sausage-like rods seen in the6-Å map actually were α-helices. There was a strongpresumption of this at 6 Å, but no proof. Moreover,the connections between “sausages” in the 6-Å mapwere unclear in places, especially where theyappeared to navigate a sharp corner.Calculation of the three-dimensional electron

density map of myoglobin, involving 9600 reflec-tions, was an all-night affair at the EDSAC IIcomputer. John Kendrew, Bror Strandberg, DavidDavies, and I, as well as our London collaboratorDavid Phillips and some of his coworkers, were allon hand. Ninety-six sections were calculatedthrough the crystals, with each section being a gridof density values spaced roughly 2/3 Å apart.15 Thecomplete electron density map required 12 h ofcomputer time. In contrast, calculation of the earlier6-Å map with 400 X-ray reflections had taken only72 min on EDSAC I.11

As the night dragged on and as completed elec-tron density sections began appearing, the sectionswere scanned for information that the low-resolu-tion map did not have. The haem group was easilyidentified (Figures 2.8 and 2.9) and, to everyone'ssatisfaction, the 6-Å “sausages” did indeed turn outto be α-helices. Side chains were clearly visible, andeven CfO groups within the helix were clear,thereby establishing the direction of the backbonechain (Fig. 2.10). The overall structure of the sperm-whale myoglobin molecule is shown in Fig. 2.11.To celebrate the new structure, a garden party

was held at dusk on the lawn of Peterhouse. Con-toured Plexiglas sheets of the 2-Å myoglobin mapwere stacked on a large light box that was the centerof attraction. One of my fondest memories of thatparty is that of Sir Lawrence Bragg, who hadpushed for protein structure analysis for more thana quarter of a century, dragging one party attendeeafter another over to the light box, pointing downthe E-helix, and saying in exultation, “Look! It'shollow! It's hollow!”Shortly thereafter, my wife Lola, my infant son

Ian (a Cambridge citizen by birth), and I dasheddown to Southampton to catch a ship home so Icould take upmy duties as a new faculty member atthe University of Illinois. Bror and David Davies

remained some months longer, helping John con-struct a display of electron density built of coloredclips on a forest of steel rods (Fig. 2.12), into whichone could build an accurate atomic structure usingwire models invented by John Kendrew. (Kendrewmodel parts are now a collector's item.) The modelwas built in the basement of the CavendishLaboratory and was put on proud display thefollowing autumn, when the International Unionof Crystallography held its biennial congress inCambridge. For the congress, the new myoglobinand haemoglobin results were displayed in a room

Fig. 2.11. An accurate but simplified diagram of the folding of α-helices in sperm-whale myoglobin from the 2-Åanalysis. Each black circle along a helix marks the α-carbon of an amino acid. Zigzag lines follow the chain backbone innonhelical regions such as bends between helices. The haem group and its attachments also are shown. Drawn by theauthor to advertise a 1962 lecture by John Kendrew at the University of Illinois.5


that belonged to the Electrical Engineering Depart-ment. Permission for use of the room was granted,but John and Max were told in no uncertainterms to get “all of that stuff” out of the roomquickly before the fall term began. The models nowreside in the Kensington Science Museum inLondon.Not long after the low-resolution 6-Å myoglobin

structure was published, someone in HowardSchachman's laboratory at Berkeley drew thecartoon seen in Fig. 2.13. John Kendrew standsuncertainly to the right of the monster that he hascreated, with computer output strewn on the floor.Schachman sent the drawing to me in 1965, butcould not recall just which of his students hadbeen the artist. The molecule clearly has the sameoverall structure as Fig. 1.4 of Chapter 1, completewith the breaks between helices and the superfluousconnections that were relics of the low-resolutionstructure.

Afterthoughts

John Kendrew and Max Perutz had quite differentstyles in running a research laboratory. John was thementor, guide, and organizer. He was simulta-neously involved in several other activities outsidethe laboratory: renovation of the art in Peterhouseand renovation of the electrical system that (as Iremember) Rutherford had first installed, scienceadvisor to the British government, liaison with theUS government over the Polaris missile program,and other useful activities. He would come into thelaboratory for a groupmeeting several mornings perweek andwould ask us, “How is progress?What hasbeen happening?What do you plan to do next?Whatdo you need?” He was valuable to the group as aresource and inspiration. In a sense, he was an ex-cellent mentor for someone who aspired running anindependent research group some day. We learnedhow to run our own show. Myoglobin, however,

Fig. 2.12. The 2-Å electron density displayed as colored clips in a forest of steel rods mounted on a plywood base. Thedetailed backbone of the polypeptide chain can be seen snaking through this forest. John Kendrew (center left) explainsmyoglobin features to attendees at the 1960 meeting of the International Union of Crystallographers in Cambridge.5


marked the end of John's scientific research career. In1974, after the conclusion of the final 1.5-Å-resolutionanalysis of myoglobin by Herman Watson andcoworkers, John left Cambridge to become the firstdirector of the European Molecular Biology Organi-zation Laboratory in Heidelberg. By all accounts, he

Fig. 2.13. Dr. Frankendrew's monster. The 6-Å model,caricatured around 1958 by a now-forgotten member ofHoward Schachman's laboratory at UC Berkeley. Thecaricature was sent to me by Schachman in 1965. Note thebreaks and false connections between helices that werecleared up by the 2-Å analysis.5

did a superb job. When he retired from there in 1981,he accepted a position, appropriately enough, aspresident or master of St. John's College at Oxford.(This is why all his files eventually wound up atOxford instead of Cambridge.) I can fantasize someawe-stricken freshman coming up to the distin-guished, white-haired President and asking timidly,“Are you the St. John forwhomthe college is named?”In contrast, Max was a hands-on bench biochemist

whose center of gravity was always the laboratoryitself. On many occasions, he would involve himselfdirectly in the ongoing structure analysis as, forexample, when he came back to the laboratory in theevenings to mount crystals for Roger Hart's next-day precession camera surveys. He became directorof the MRC Laboratory and served for many years,but this did not stop his daily involvement with thelaboratory. Both styles had their merits: one learnedfrom John, but one learned with Max. Max's lifelongcommitment to research is acknowledged in Fig.2.14, which depicts him as the modern Noah of theMRC, bringing proteins safely through the delugeon an ark that he himself constructed. The app-roaching dove at right carries the reward that Maxand John earned for their efforts.In 1996, the International Union of Crystallogra-

phy held its congress in Seattle, WA. A panel ofcrystallographic Nobel laureates was invited tospeak about their life in structure analysis. Onequestion asked by the moderator was: “If you had tolive your life over again, would you do again justwhat you actually did?” My graduate mentor BillLipscomb and most other Nobel laureates replied,“Of course.” But John Kendrew had a different takeon things: over the years, he had come to appre-ciate how much he enjoyed scientific administration

Fig. 2.14. Max as Noah, bring-ing his proteins safely through thedeluge. Drawn by editorial cartoo-nist Dave Granlund after myrough sketches for the cover ofPresent at the Flood: How StructuralMolecular Biology Came About.5Note, to be rigorously fair, thatthe myoglobin molecule leadingoff at the prow of the Ark moreproperly belongs to John Kendrew,not to Max. But it was indeed first.For more superb cartoons byGranlund, see his Web site [www.davegranlund.com/cartoons].


at the European Molecular Biology OrganizationLaboratory. If he had to do it over, he said, heprobably would not have spent so many years onresearch, but would have moved into administra-tion much earlier. But there is an interesting flaw inthis argument. He was presumably offered theEuropean Molecular Biology Organization positionbecause he was both a capable administrator and aNobel laureate. Had he not solved the myoglobinstructure or carried out some comparable notablescientific achievement, it was unlikely that he wouldhave been chosen as the founding leader of theEuropean Molecular Biology Organization Labora-tory. But all things turned out for the best for bothJohn and Max. De mortuis nil nisi bonum.

References

1. Dickerson, R. E., Wheatley, P. J., Howell, P. A. &Lipscomb, W. N. (1956). Boron arrangement in a B9hydride. J. Chem. Phys. 25, 606–607.

2. Dickerson, R. E., Wheatley, P. J., Howell, P. A. &Lipscomb, W. N. (1957). Crystal and molecular struc-ture of B9H15. J. Chem. Phys. 27, 200–209.

3. Dickerson, R. E. & Lipscomb, W. N. (1957). Semitopo-logical approach to boron-hydride structures. J. Chem.Phys. 27, 212–217.

4. Dickerson, R. E. (1992). A little ancient history. ProteinSci. 1, 182–186 (Reproduced in Reference 5).

5. Dickerson, R. E. (2005). Present at the Flood: How Struc-tural Molecular Biology Came About. Sinauer Associates,Sunderland, MA.

6. Kendrew, J. C. & Parrish, R. G. (1956). The crystalstructure of myoglobin: III. Sperm-whale myoglobin.Proc. R. Soc. London Ser. A, 238, 305–324.

7. Bennett, J. M. & Kendrew, J. C. (1952). The compu-tation of Fourier syntheses with a digital electroniccalculating machine. Acta Crystallogr. 5, 109–116.

8. Green, D. W., Ingram, M. & Perutz, M. F. (1954). Thestructure of haemoglobin: IV. Sign determination bythe isomorphous replacement method. Proc. R. Soc.London Ser. A, 225, 287–307.

9. Harker, D. (1956). The determination of the phases ofthe structure factors of non-centrosymmetric crystalsby the method of double isomorphous replacement.Acta Crystallogr. 9, 1–9.

10. Kendrew, J. C., Bodo, G., Dintzis, H. M., Parrish, R. G.,Wyckoff, H. & Phillips, D. C. (1958). A three-dimen-sional model of the myoglobin molecule obtained byX-ray analysis.Nature (London), 181, 662–666 (in Ref. 5).

11. Bodo, G., Dintzis, H. M., Kendrew, J. C. & Wyckoff,H. W. (1959). The crystal structure of myoglobin: V. Alow-resolution three-dimensional Fourier synthesis ofsperm-whale myoglobin crystals. Proc. R. Soc. LondonSer. A, 253, 70–102.

12. Dickerson, R. E., Kendrew, J. C. & Strandberg, B. E.(1961). The crystal structure of myoglobin: phase deter-mination to a resolution of 2 Å by the method of iso-morphous replacement. Acta Crystallogr. 14, 1188–1195.

13. Dickerson, R. E., Kendrew, J. C. & Strandberg, B. E.(1961). The phase problem and isomorphous replace-ment methods in protein structures. ComputingMethods and the Phase Problem in X-ray Crystal Analysis.pp. 236–251, Pergamon Press, New York.

14. Blow, D. M. & Crick, F. H. C. (1959). The treatment oferrors in the isomorphous replacement method. ActaCrystallogr. 12, 794–802.




15. Kendrew, J. C., Dickerson, R. E., Strandberg, B. E.,Hart, R. G., Davies, D. R., Phillips, D. C. & Shore, V. C.(1960). Structure of myoglobin: a three-dimensionalFourier synthesis at 2 Å resolution. Nature (London),185, 422–427 (in Ref. 5).

Chapter 3: Recollection of the EventsLeading to the Discovery of theStructure of haemoglobin

Michael G. Rossmann

Fig. 3.1. TheMedical Research Council Hut at the NewMuseum Site off Free School Lane, Cambridge, photo-graphed in 2003, still almost unchanged after nearly50 years. However, in 2003, it was all locked up, whereasin 1959, it was full of activity day and night.

Introduction

My PhD work in the laboratory of MonteathRobertson at the University of Glasgow in Scotlandrequired lengthy calculations of Fourier seriessummations to determine electron densities. Ourcomputational equipment consisted of desk calcula-tors powered by electric motors and Beevers–Lipsonstrips representing sine and cosine curves withdifferent amplitudes. It used to take me aboutone week to calculate one two-dimensional projec-tion for the flat aromatic molecules I was analyzing.It was, therefore, a great delight for me to encounterat the University of Minnesota inMinneapolis one ofthe first commercial computers, the UNIVAC 1103,when I joined the laboratory of Bill Lipscomb (the“Colonel”) as a young postdoctoral fellow. TheColonel had arranged for a few hours per week ofcomputing time for his group on the UNIVACcomputer at Honeywell in neighboring St. Paul. Ioverlapped with Dick Dickerson for a couple ofmonths as he finished his doctorate on the structureof boron hydrides, a subject that gained the NobelPrize for the Colonel about 15 years later. Dick taughtme my first lessons on how to program a computer.Before I left Minnesota two years later, Bob Jacobson,a graduate student with the Colonel, and I hadwritten a fairly complete crystallographic computingpackage that continued to be in use for many years.While in Minnesota, I attended the International

Union of Crystallography meeting in Montrealand listened to a fascinating lecture by DorothyHodgkin1 on the structure of myoglobin beingtackled by John Kendrew in Cambridge. Work onthe structure determination of proteins seemed, tome, to be the best kind of project for my interests inmethods for solving the crystallographic phaseproblem and the use of modern computational tech-niques. Fortunately, Max Perutz, the director of theCambridge Laboratory that also included JohnKendrew, quickly replied positively to my requestfor a job.When my family and I arrived in Cambridge, I

found a kind of environment different from what Ihad known either in Glasgow or Minneapolis. Weworked in the little “Hut” (still standing even today;Fig. 3.1) in the New Museum Site off Free SchoolLane, close to the central market square. I gathered

that there was insufficient space in the CavendishLaboratory, which was world famous for the workof Thompson, Rutherford, and others on thestructure of the atom, to also accommodate MaxPerutz and members of his group who were study-ing biological molecules. Thus, we were all crowdedinto The Hut, next to the Austin Wing of theCavendish. I shared a very small office with BrorStrandberg, a Swedish graduate student; DickDickerson, who had arrived via a circuitous routeafter a short stay at the University of Leeds; andLarry Steinrauf, a former student of Linus Pauling.The office opposite to ours housed Francis Crick andSydney Brenner. Max and John each had their ownoffice. There was one office for the “computer girls”who helped with manual calculations, another officefor Ann Cullis (assistant to Max Perutz) and MaryPinkerton (assistant to John Kendrew), an office for agraduate student (Hilary Muirhead) and assortedpostdoctoral fellows who worked for Sydney andFrancis, plus a very crowded biochemistry labo-ratory. Everyone packed tightly into the centralcorridor of The Hut at 11 o'clock each morning,reaching for the coffee pot brewed by Leslie Barnett,Sydney Brenner's assistant, while debating every-thing from the relation between church and sciencein the then new Churchill College to the best ways offinding heavy-atom positions in protein crystals.Many a morning the conversations were dominatedby Francis's loud but appealing voice and provoca-tive ideas. This was, for me, an intellectual environ-ment more stimulating than I had ever previouslyencountered. I started to realize that biology was amore interesting subject than I had thought and torecognize how ignorant I had remained inmy narrowundergraduate physics and mathematics education.No longer would the phase problem be my onlychallenge, as I became aware that the most signi-ficant problems were those of the way the universeand Earth had evolved to make the present day.


Protein Structure in Cambridge before1958

The likelihood of ever solving the three-dimen-sional structure of a protein seemed extremely smalleven in the 1950s. There were too many data to becollected, no reasonable way of solving the phaseproblem, and no way of computing the results tomake an interpretation of the collected data. Butmaybe even more daunting was the nagging ques-tion of whether each protein molecule had a uniquestructure. To some extent, these problems werestarting to be solved. Bernal and Dorothy Crowfoot(later Hodgkin) had shown that pepsin could beinduced to grow crystals that could diffract to near-atomic resolution,2 implying that all the moleculesin the crystal had the same structure. Max encour-aged Tony Broad to build a powerful rotating-anodeX-ray generator. Max had also overseen the devel-opment of an accurate instrument for measuring theintensity differences between the X-ray reflectionsfrom isomorphous heavy-atom derivatives of hae-moglobin crystals. Punched-card-driven devices,which were first used in the English textile industryand later became standard items of office equipmentas the forerunners of computers, were starting to beused for mathematical computations. Furthermore,Max had been creating a set of very original tech-niques for solving the phase problem. One of thesewas the isomorphous replacement technique thathad been the basis of the first crystal structures eversolved by X-ray diffraction, namely, those of NaCland KCl.3 It had also been explored by my formermentor Monteath Robertson as a tool for determin-ing the structure of some haem-like compounds,4

but most crystallographers at the time thought thatthe incorporation of even the heaviest atoms wouldhave only a negligible impact on the X-ray diffrac-tion of a protein crystal. Nevertheless, in 1953, Maxhad shown that, using heavy-metal isomorphousreplacements, it was possible to determine thephases of the (h0l) reflections of his monoclinichaemoglobin crystals.5 Admittedly, these reflectionswere of a centric zone and therefore limited to beeither 0° or 180°, but this was enough to give hopethat a structure determination might be possible.John Kendrew, a former student of Max, had

initiated a study of the smaller myoglobin molecule.Myoglobin is a carrier of oxygen in muscles, muchlike haemoglobin is a carrier of oxygen in blood.However, myoglobin has a molecular weight of only17 kDa and consists of a single polypeptide chain,whereas haemoglobin consists of four polypeptidechains: two α- and two β-chains each of about17 kDa molecular weight. With his smaller proteinmolecule, John had made more rapid advances thanMax, using the techniques Max had developed. In1957, John had been able to obtain a 6-Å-resolutionelectron density map of sperm-whale myoglobinthat seemed to represent a series of bent helices,6

probably α-helices as had been proposed by LinusPauling.7 Models of John's myoglobin structure in a

variety of different scales were everywhere in TheHut. This structure had been the big event in theyear before I arrived in Cambridge and had beenheralded as being imminent by Dorothy when Iheard her speak in Montreal.

Autumn of 1958

By the time I arrived in Cambridge in September1958, Max and Ann Cullis had regained their abilityto grow good crystals of monoclinic oxygenatedhaemoglobin and were preparing a series of heavy-atom derivatives based on the substitution of acouple of Cys amino acids with Hg derivatives. Maxhad determined the x and z coordinates using the(h0l) projection data he had so carefully exploredover the previous six years or more. The unresolvedproblem was the determination of the relative ycoordinates of the heavy-atom sites. This problemexisted because there is no centric projection perpen-dicular to the unique 2-fold axis in a monoclinicspace group and no defined origin as is provided bythe crystallographic 2-fold axis in the (h0l) pro-jection. John Kendrew had encountered this pro-blem in the determination of the 6-Å-resolutionmap of myoglobin. Perutz8 and Blow9 had madesuggestions as to how this problem might be solved,but none of the proposed functions gave unambig-uous solutions.A few steps from The Hut was an old, ugly, brick

building that housed the very new MathematicsLaboratory. In this building, Maurice Wilkes and his“boffin” friends had put together a machine fromleftover wartime radar equipment that was cap-able of being programmed to do all manner of fastcomputations. This Electronic Delay Storage Auto-matic Calculator (EDSAC) was arguably the firstautomatic stored-program electronic computer everbuilt. John Kendrew had used this machine tocalculate the 6-Å-resolution map of myoglobin.However, by the time I arrived in Cambridge, theoriginal EDSAC had been replaced by a new, muchfaster, muchmore powerful EDSAC II computer thathad just opened its doors to other selected users(meaning friends and acquaintances of college“dons”). EDSAC II was the brainchild of DavidWheeler. It was quickly apparent to me that EDSACII was far more sophisticated than the machine I hadused in Minnesota. I read the well-written manualand started putting some of my crystallographicexperience into this wonderful new machine. How-ever, I did not have a Cambridge background andwas not aware of the conventions assumed by in-group users. As a result, I was treated with somesuspicion that was not dispelled until I was able toshow that my programs produced correct results inan acceptably short time frame. In addition, therewas a large group of more conventional crystal-lographers at Cambridge who distrusted the newcomputers, while others, likemyself, advocated theiruse. The prevailing thought was that the humanmind should be more powerful than any machine.


One of the first tasks that I set for myself was towrite a Fourier synthesis program for the calculationof electron density and other crystallographic maps.I also proposed a new and simpler function to solvethe “y-axis” problem mentioned above. Althoughprogram testing could be performed at certainspecific times during the day, large-scale computa-tions could be performed only on allocated nights.Max and John had been given Monday nights, to beshared with the radio astronomers of Martin Ryle'sgroup. Thus, one Monday, maybe in November1958, I used my new Fourier program to computemy new y-axis correlation functions. Alas, when Icame to examine the results the next morning, Irealized that the symmetry was wrong. Clearly,there was a mistake (maybe many) in the program.Max and Ann were very supportive and assured methat they could wait for me to try to fix my program.Fortunately, the problem was not serious and,two weeks later, I was back on another Mondayevening repeating the same calculations with acorrected program.My delight was indeed great the next morning.

The maps showed exactly what I had hoped—positive peaks for vectors between atoms in the samecompound and negative peaks for vectors betweenatoms in different compounds10 (Fig. 3.2). Further-more, the vectors were all self-consistent. Best of all,the peakswere sky high, leaving absolutely no doubtin differentiating between signal and noise. Now itwas certain that we had the necessary heavy-atomderivatives, the data-processing procedures, and the

Fig. 3.2. Composite view of the (FHgAc2−FPCMB)2

correlation function between the HgAc2 and para-chlor-omercurybenzoate (PCMB) heavy-atom derivatives. Con-tinuous lines are positive contours, whereas broken linesare negative contours. The Harker sections at v=0 andv=1/2 have positive peaks relating the symmetry-relatedHgAc2 sites and also the symmetry-related PCMB sites,whereas the v=0.3 section contains negative vectorsbetween the HgAc2 and the PCMB sites.10

technology to find the correct positions of the heavyatoms, all prerequisites for determining the structureof haemoglobin. Max was delighted. He brought inevery person who would understand the signifi-cance of these results, starting with the local crystal-lographers such as Bill Cochran and then the out-siders such as Sir Lawrence Bragg, Dorothy, andBernal (the “Sage”). Mostly, he would ask me for myfolders containing the maps whenever he had a dis-tinguished visitor, but it was obvious that he washappy to have had this progress. At the same time, Ibegan to be recognized by thosewho had complainedabout my unconventional ways and techniques.

Winter and Spring of 1959

Now we had to start thinking about phase deter-mination given the heavy-atom sites. But we stillneeded to determine the parameters that definedthe size of heavy-atom substitutions (occupanciesand “temperature” factors). For this, I wrote a leastsquares program based on the y-axis correlationmethod.John Kendrew and Sir Lawrence Bragg had used a

manual technique for determining the phases ofeach reflection based on David Harker's diagram.11

Each of the about 400 reflections used in the 6-Å-resolution map of myoglobin had been analyzedwith a “Harker diagram.” This was, in reality, anArgand diagram on which were drawn calculatedvectors representing the structure factors of theheavy atoms given their parameters. Circles corres-ponding to the observed amplitudes of all theisomorphous compounds were then drawn. Wherethe circles intersected was the most probable phase.John and Sir Lawrence had simply guessed at whatthey thought was the best point of intersection of thephase circles and averaged their guesses. Now,however, we had five times as many reflections asJohn had had for his earlier work.David Blow had been a graduate student of Max,

although in 1958, he was in America as a post-doctoral fellow with Alex Rich, first at the NationalInstitutes of Health and later at the MassachusettsInstitute of Technology. David had spent his PhDtime thinking with Francis about how to be morerigorous12 in the selection of the best phase from theHarker diagram. Dick and Bror simplified the Blow–Crick ideas to a readily programmable algorithm.13 Ihad helped to teach Hilary Muirhead to programEDSAC II and suggested that she should write aphasing program. I was, however, not entirelyhappy with Dick and Bror's algorithm because itassumed that all the phasing errors were in themeasurement of heavy-atom derivative data, withno error in unsubstituted “native” data. Thus,Hilary and I somewhat modified the procedure14—a change that most probably aided in producing theclean map that we subsequently calculated inAugust 1959.Max had briefly taught me how to mount the

haemoglobin crystals into glass capillaries, how to


operate the Broad X-ray generators in the basementof the Austin Wing of the Cavendish Laboratory,and how to handle precession cameras withprotein crystals. Among the man-made obstacleswas the chief Cavendish Laboratory technicianwho could never understand why we needed somuch film to collect our data. Persuading him toprovide a new box of film was a major diplomaticfeat that usually resulted in being given merely afew films at a time.We now had all the tools and all the data I thought

were necessary for producing a 5.5-Å-resolutionmap of haemoglobin. Max and Ann had collectedabout five heavy-atom derivative data sets, but Maxwanted to collect one more. Although, in retrospect,that can now be seen to have been an unnecessaryprecaution, we had no idea what we would see inthe map. Thus, when we eventually saw the electrondensity, there was no doubt about its significanceand, considering the unexpected nature of the result,it might have been met with disbelief had the mapbeen poorer.

August 1959

Everything was ready by August and Max agreedto the calculation of the haemoglobin electrondensity map. That is everything except for EDSACII, which had had one of its frequent nervousbreakdowns. We impatiently waited another

Fig. 3.3. The 5.5-Å-resolution haemoglobin map. Section y

two weeks when Max was invited to be the firstuser of the repaired computer. I persuaded Max tocome with me to the Math lab that afternoon. I triedto teach Max how to feed the punched paper tapesinto the computer, but he was too nervous, soeventually I started the computations myself. Aboutan hour and a half later, the map was calculated, butit still needed plotting by hand, a tedious task thatwould consume our computer girls for the next fewdays. But I did notice that there were two sectionswith very high density, which presumably were thehaem groups in the α- and β-chains of the molecule.(The space group was C2, with molecules sitting oncrystallographic 2-fold axes.) Thus, I suggested thatwe should explore these positions with a finer gridinterval. Max agreed, and I punched a couple ofquick “jiffy tapes” to do this task. This time, Maxwas less nervous and succeeded. Unfortunately,seconds after Max had initiated the calculations,smoke started to pour out from behind the controlpanel, and everything closed down. Max made acomment, something like, “You see, I should nothave operated the computer.” EDSAC II was closedfor another two weeks. Then John Kendrew, DickDickerson, Bror Strandberg, and David Phillips(their collaborator from the Royal Institution inLondon) had their turn in using EDSAC II tocalculate the 2-Å-resolution map of myoglobin. Iremember receiving a phone call late that eveningfrom David Phillips asking me to come to the Mathlab to sort out a problem with my Fourier program.

=1 showing long rods of density representing α-helices.


They then took most of the night to finish theircalculation.Once the haemoglobin maps had been plotted

onto a semitransparent tracing paper (Figs. 3.3–3.5),Max tried to construct a cage made of Meccano partsin his office in order to display the map sections in away that might help make the map more easilyinterpretable. Yet when I looked at the maps, I couldsee long, thin features that were surely α-helices. Ireally wanted to look at the maps and not wait onMax. However, maybe out of force of habit, Maxstarted to prepare more heavy-atom derivatives inthe laboratory. What happened next I kept to myselffor many decades. I had been privileged to work onthe haemoglobin project with Max, but it was also aproject that Max had given his whole life to develop.In my enthusiasm to look at the results, I stole thefinal discovery from Max. When I realized what Ihad done, I felt terrible. I have never truly forgivenmyself for my behavior. The excitement of discoverywas completely shattered for me.When Georgina Ferry was doing her research

prior to writing Max's biography, she came tointerview me at Purdue. I felt it right to tell hereverything I knew about Max, which would have toinclude these moments on the summit of MountDiscovery as a background to understanding thecharacter of her biographical study. Sometime afterher visit, she sent me her presentation of thoseevents with a request that she be allowed to use thispassage in her upcoming book.15 I found her writing

Fig. 3.4. The 5.5-Å-resolution haemoglobin map. Section y

to be sensitive, balanced, and accurate; thus, I gaveher permission to use this passage in her book. Forme, this was also a catharsis on a topic about which Ihad spent much anguish.The helices in the maps were all connected

by density permitting a trace of both the α and βpolypeptide chains. I had once contemplated, longbefore the map was out, that there might be somesimilarity between myoglobin and haemoglobinbecause of the coincidences of molecular weightsand functions. But I was quickly put down withdisdain. Nevertheless, I was not overly cowed bythese comments because, having traced the densitychains, I immediately tried to compare the easilyavailable three-dimensional 6-Å-resolution myoglo-bin model with my chain tracing. The result wasobvious.I found Max in the Biochemistry Laboratory. Only

then did I realize how much I had hurt him. Withthe realization of what I had done, all desire toexplore further was completely gone. Instead, Maxused his outstanding experimental skills to build amodel from thermally setting clay (Fig. 3.6), which Iwould never have had the skill or imagination to domyself. This was a perfect way of representing thestructure, showing not only the “globin fold” butalso indicating the level of error and the packing ofhelices. The model was beautiful! Nature wasbeautiful!Notwithstanding the similarity of the myoglobin

structure with the α- and β-chains of haemoglobin,

=9 showing long rods of density representing α-helices.

Fig. 3.5. Interpretation of the 5.5-Å-resolution haemoglobin map. I used different colors to trace regions of continuousdensity. A change in color represented uncertainty of the connectivity of densities between the colored traced sections.Numbers indicate the sections onwhich density was found. I could recognize the two independent, fully connected chainsonly when this tracing had been completed. In 2004 or 2005, I gave this map, plus my model of haemoglobin, forsafekeeping to the Medical Research Council Laboratory of Molecular Biology historical archives.


Max still felt it necessary to make further checks onthe correctness of the structure, in particular of thesite of the all-important heme groups. He noted thatthe Fe atoms in the center of the heme groupsshould give rise to a measurable anomalous X-raydiffraction effect. Thus, given the now-establishedFe atom positions and the isomorphous replace-ment determination of the “native” phases, itwould be possible to calculate the reflections thatshould have the largest anomalous effect and thesign of their Bijovet difference. In this way, Idetermined 32 reflections that should have thelargest anomalous difference. Max then asked TonyNorth at the Royal Institution to measure the sizeand sign of the Bijovet difference for these reflec-tions using the diffractometer that Uli Arndt andDavid Phillips had designed to help John Kendrewwith the myoglobin data collection. Tony foundthat his experimental measurements agreed withmy predictions for 28 of 32 listed reflections.16 Thisconfirmation also had a further benefit, namely, indetermining the absolute hand of the haemoglobinmap. That information could then be applied toestablish the absolute hand of the 2-Å-resolutionmyoglobin map that Kendrew's group had justdetermined using the reasonable assumption thatthe myoglobin and haemoglobin structures have

the same hand. This demonstrated that all the α-helices in the myoglobin map were right handed.This was also a satisfactory result, as there wouldbe fewer steric clashes of the side chains oflaevorotatory amino acids with main-chain atomsin a right-handed α-helix, given the correctness ofBijovet's determination of the absolute hand ofamino acids being laevorotatory.17

The evolutionary significance of the haemoglobinstructure16,18 was not lost on anyone. Here was averification of Darwin's ideas at the molecular—andtherefore genetic—level. Francis Crick and JimWatson had suggested the manner in which a genedetermined the amino acid sequence of a protein,although the actual genetic code was still unknownat that time. Yet it would seem to be likely that theremust have been an ancestral gene that coded for aprimordial globin fold from which the present-daysperm-whale myoglobin and horse haemoglobin α-and β-chains had evolved by gene duplication andmutation. For me, this discovery has had moreinfluence on my subsequent scientific direction thanany other single event.It was a fortuitous coincidence that, in the follow-

ing year, the International Union of Crystallographyhad scheduled its triennial meeting in Cambridge. Ithad been three years since I had heard Dorothy talk

Fig. 3.6. Max's model of the 5.5-Å-resolution map ofhaemoglobin.18 A thermally setting clay was molded byMax to correspond to the density on each section outlinedby a chosen contour level. The different sections were thenglued together. The photograph of the model is onlyavailable in black andwhite, but the α-chain was shown inwhite, the β-chain was shown in black, and the hemeswere shown in red. The model referred to in the caption toFig. 3.5 was made of wood, using a thread saw to cutalong the contour lines, thereby isolating the higher-density regions that were subsequently glued together.The scale was the same as what Max had used to constructhis model and the same as the original maps shown inFigs. 3.3–3.5.


at the previous International Union of Crystal-lography meeting in Montreal. What tumultuousyears those turned out to be! The 1960 meeting inCambridge was an opportunity to display our newdiscoveries. Max gave a great lecture in the localmovie theater located in the market square, the onlyhall big enough to house the whole meeting. (Inthose days, movie theaters were bigger, andscientific conferences were smaller.)

Fig. 3.7. Dick and Bror with the myoglobin sortingboard. Each piece of punched tape represented a set ofintensities from one line (constant h and k) of reflections onone precession photograph. The same line of reflectionsmight have been recorded on a number of photographs,thus permitting the determination of scale factors betweenphotographs.

After August 1959

Not long after those extraordinary days in 1959,my three officemates dispersed around the world.Dick and Larry returned to America, and Brorstayed on a few months to initiate the building ofthe myoglobin model in the cold basement of theAustin Wing before returning to Sweden. But thedesk vacated by Larry was quickly occupied byDavid Blow on his return from the United States.Although he had missed two exciting years, he(as was I) was fascinated by the crystallographicimplications of structure determinations that hadhappened in his absence. One of David's interests asa graduate student had been the use of anomalousdispersion to aid phase determination. I had alsobeen thinking about the use of anomalous dis-persion for finding the position of anomalousscatterers.19 We found ourselves in daily stimu-lating discussions, producing, in the subsequent

five years, a series of papers that have helped toform the foundations of modern structural biology.The topics we covered were single isomorphousreplacement (SIR) and single anomalous disper-sion (SID),20 phase combination,21 and molecularreplacement (MR).22–25 The genesis of these paperscame largely from the haemoglobin structuredetermination. We asked ourselves how few deri-vatives would have really been necessary to solvethe structure (SIR and SAD); how crystallographicstructure factor calculations should be compared tomultiple isomorphous replacement phasing; andwhether it would have been possible to discoverthe relationship between the α- and β-chains ofhaemoglobin without actually determining thestructure (MR).David initiated a study on chymotrypsin, a study

that we initially shared equally.26 Not only werewe able to use all our new ideas, but the dataprocessing also benefited from our collaboration.Whereas Dick and Bror had used a pseudo-manual/computer sorting procedure captured ona now-famous photograph (Fig. 3.7), I developed acomputer sorting system for the chymotrypsinstudies that I later adapted and improved for useat Purdue University and in modern commercialsoftware.27

Max made it clear to me that I should try to find amore permanent job somewhere else. Fortunately,there was no shortage of job offers once we haddetermined the structure of haemoglobin, but theywere all from America. I accepted the offer fromPurdue University because I felt that the head of theBiology Department was genuinely interested in


helping me to work on the structure of proteins. Mydecision turned out to be right. Although for manyyears I missed the intense intellectual stimulus ofCambridge, challenges at Purdue allowed me to puta personal imprint on the development of myown laboratory, which might not have been possiblein a more established environment. I remain enor-mously grateful to Max, Dorothy, David Phillips,and many other English colleagues who stronglysupported my work at Purdue in numerous ways,including frequent visits and nominations to pres-tigious awards. It is significant that exactly 50 yearssince the discovery of the haemoglobin structure,we are about to move into the completely new“Hockmeyer Hall of Structural Biology” at PurdueUniversity.

Acknowledgements

I thank Georgina Ferry for reading and correctingmy original draft of this manuscript, and SherylKelly for help in its preparation for publication.1959 was both a year of joyful discovery and a

year of domestic problems for me. I remain enor-mously grateful to Lola and Dick Dickerson fortheir extremely kind and generous help withoutwhich I would not have been able to do the workdescribed in this article. Similarly, the concern ofMax and Ann (Cullis) Kennedy for my welfareremains an overriding component of my memories.Although Max was not initially a fan of computer-ized technology, he was always willing to searchfor ways to satisfy my appetite for what I deemednecessary within reasonable financial limits. I alsothank David Blow for the fun years we hadtogether, as well as my two “computer girls” Jill(Collard) Dawes and Angela (Campbell) Mott fortheir outstanding technical help and lifelongfriendship. Finally and most importantly, I thankmy wife Audrey for her faithful loving support inthe face of strong competition from science, myother love.

References

1. Ferry, G. (1998).Dorothy Hodgkin: A Life.Granta Books,London.

2. Bernal, J. D. & Crowfoot, D. (1934). X-ray photographsof crystalline pepsin. Nature (London), 133, 794–795.

3. Bragg, W. L. (1913). The structure of some crystals asindicated by their diffraction of X-rays. Proc. R. Soc.London Ser. A, 89, 248–277.

4. Robertson, J. M. (1935). An X-ray study of thestructure of phthalocyanines: Part I. The metal-free,nickel, copper, and platinum compounds. J. Chem. Soc.615–621; 1935.

5. Green, D. W., Ingram, M. & Perutz, M. F. (1954). Thestructure of haemoglobin: IV. Sign determination bythe isomorphous replacement method. Proc. R. Soc.London Ser. A, 225, 287–307.

6. Bodo, G., Dintzis, H. M., Kendrew, J. C. & Wyckoff,

H. W. (1959). The crystal structure of myoglobin: V. Alow-resolution three-dimensional Fourier synthesis ofsperm-whale myoglobin crystals. Proc. R. Soc. LondonSer. A, 253, 70–102.

7. Pauling, L., Corey, R. B. & Branson, H. R. (1951). Thestructure of proteins: two hydrogen-bonded helical con-figurations of the polypeptide chain. Proc. Natl Acad.Sci. USA, 37, 205–211.

8. Perutz, M. F. (1956). Isomorphous replacement andphase determination in non-centrosymmetric spacegroups. Acta Crystallogr. 9, 867–873.

9. Blow, D. M. (1958). The structure of haemoglobin:VII. Determination of phase angles in the non-centrosymmetric [100] zone. Proc. R. Soc. London Ser.A, 247, 302–336.

10. Rossmann, M. G. (1960). The accurate determinationof the position and shape of heavy-atom replacementgroups in proteins. Acta Crystallogr. 13, 221–226.

11. Harker, D. (1956). The determination of the phases ofthe structure factors of non-centrosymmetric crystalsby the method of double isomorphous replacement.Acta Crystallogr. 9, 1–9.

12. Blow, D. M. & Crick, F. H. C. (1959). The treatment oferrors in the isomorphous replacement method. ActaCrystallogr. 12, 794–802.

13. Dickerson, R. E., Kendrew, J. C. & Strandberg, B. E.(1961). The crystal structure of myoglobin: phase deter-mination to a resolution of 2 Å by the method of iso-morphous replacement. Acta Crystallogr. 14, 1188–1195.

14. Cullis, A. F., Muirhead, H., Perutz, M. F., Rossmann,M. G. & North, A. C. T. (1962). The structure of hae-moglobin: VIII. A three-dimensional Fourier synthesisat 5.5 Å resolution: determination of the phase angles.Proc. R. Soc. London Ser. A, 265, 15–38.

15. Ferry, G. (2007). Max Perutz and the Secret of Life.Chatto and Windus, London.

16. Cullis, A. F., Muirhead, H., Perutz, M. F., Rossmann,M. G. & North, A. C. T. (1962). The structure ofhaemoglobin: IX. A three-dimensional Fourier synth-esis at 5.5 Å resolution: description of the structure.Proc. R. Soc. London Ser. A, 265, 161–187.

17. Bijvoet, J. M., Peerdeman, A. F. & van Bommel, A. J.(1951). Determination of the absolute configurationof optically active compounds by means of X-rays.Nature (London), 168, 271–272.

18. Perutz, M. F., Rossmann, M. G., Cullis, A. F.,Muirhead, H., Will, G. & North, A. C. T. (1960). Struc-ture of haemoglobin. A three-dimensional Fouriersynthesis at 5.5-Å resolution, obtained by X-rayanalysis. Nature (London), 185, 416–422.

19. Rossmann, M. G. (1961). The position of anomalousscatterers in protein crystals. Acta Crystallogr. 14,383–388.

20. Blow, D. M. & Rossmann, M. G. (1961). The singleisomorphous replacement technique. Acta Crystallogr.14, 1195–1202.

21. Rossmann, M. G. & Blow, D. M. (1961). The refinementof structures partially determined by the isomorphousreplacement method. Acta Crystallogr. 14, 641–647.

22. Rossmann, M. G. & Blow, D. M. (1962). The detectionof sub-units within the crystallographic asymmetricunit. Acta Crystallogr. 15, 24–31.

23. Rossmann, M. G. & Blow, D. M. (1963). Determinationof phases by the conditions of non-crystallographicsymmetry. Acta Crystallogr. 16, 39–45.

24. Rossmann, M. G., Blow, D. M., Harding, M. M. &Coller, E. (1964). The relative positions of independentmolecules within the same asymmetric unit. ActaCrystallogr. 17, 338–342.


25. Dodson, E., Harding, M. M., Hodgkin, D. C. &Rossmann, M. G. (1966). The crystal structure ofinsulin: III. Evidence for a 2-fold axis in rhombohedralzinc insulin. J. Mol. Biol. 16, 227–241.

26. Blow, D. M., Rossmann, M. G. & Jeffery, B. A. (1964).The arrangement of α-chymotrypsin molecules in themonoclinic crystal form. J. Mol. Biol. 8, 65–78.

27. Rossmann, M. G., Leslie, A. G. W., Abdel-Meguid,S. S. & Tsukihara, T. (1979). Processing and post-refinement of oscillation camera data. J. Appl. Crystallogr.12, 570–581.

Epilogue

John Kendrew met J. D. Bernal when they wereboth in military service in Southeast Asia during theSecond World War. Bernal predicted to John that thegreat postwar challenge would be to determine thethree-dimensional structure of a protein by means ofX-ray crystallography. Bernal's final advice to Ken-drew was: “When this is over, go home to Englandand solve the structure of a globular protein.” Later,after the myoglobin structure had been solved,Kendrew's comment to this advice was, “And so Idid.”

Perhaps neither Bernal nor John realized, even afterthe structure of myoglobin had been determined, thatthis was the very beginning of a vast scientific revo-lution. The structures of myoglobin and haemoglobin(Fig. E1) were the distant rolls of thunder, unnoticedby most, of a tremendous storm that was gatheringand that would completely change the biologicallandscape.

After that summer of 1959, nothing much hap-pened in structural biology for several years. Thegeneral opinion was that perhaps there would be nofurther protein structures any time soon. Indeed, theworld had to wait until 1967 for the structure of henegg white lysozyme, determined by David Phillips'team at the Royal Institution in London and over-

Fig. E1. John and Max standing behind the forest ofrods into which the first sperm-whale myoglobin modelwas built in the basement of the Cavendish Laboratory.Reprinted with permission from the MRC Laboratory ofMolecular Biology.

seen by Sir Lawrence Bragg.1,2 Lysozyme not onlyhad elements of β-sheets as had been predicted byPauling,3 but it was also the first enzyme whosestructure was determined. Furthermore, the structurewas able to put onto a firm basis many of the previoussuggestions on how an enzyme can catalyze achemical reaction.

Over the subsequent few years, a few other en-zymes, including carboxypeptidase, chymotrypsin,ribonuclease, papain, insulin, carbonic anhydrase,and lactate dehydrogenase, were determined. Werealized that, suddenly, there were quite a lot of datathat needed to be made accessible to all. At the ColdSpring Harbor meeting in the summer of 1971 therewas a small meeting that included Max Perutz, JanDrenth, Fred Richards, Walter Hamilton, and MichaelRossmann, which established the Protein Data Bankfor the collection of all protein (and later nucleic acid)structures. The holdings of the Protein Data Bankgrew only slowly at first, but the rate has increasedexponentially to where about a total of 50,000 struc-tures have now been deposited. The major factorsthat have made this possible are the advent of syn-chrotron radiation for X-ray diffraction studies, freez-ing of crystals to greatly reduce radiation damage,and automatic computerized data processing as wellas improved methods for phase determination andphase refinement—all making it possible to solvecrystal structures within hours of collecting the firstdiffraction data.

Biological revolution is not only the speed withwhich structures can now be solved, but also theincreasing biological significance of the structures.Structures of viruses, ribosomes (barely known toexist in 1959), and ribozymes (not dreamed of in 1959)have become available. All these varied topics havebeen found to have a common thread in their evo-lution, as it has become apparent that many biologi-cally essential functions are based on structures thathave changed little since the appearance of life onEarth. Themodern crystallographer knows little aboutspace groups or structure factors, but does know agreat deal about structure and biology. A new science,called “bioinformatics,” has arisen—made necessaryby the enormous growth of structural and genesequence information. Indeed, structural biology hasbecome the unifying factor of just about every aspectof biology.

In 1959, we were a small elite group of friends whohad a partial and incomplete vision of what might bethe future. Today, structural biologists are on everycontinent in every country. The biological knowledgeexhibited by the newest generation of scientists,whether physicist, chemist or biologist, is veryimpressive, but he/she will probably know little, ifanything, about what happened in 1959.

References

1. Blake, C. C. F., Koenig, D. F., Mair, G. A., North,A. C. T., Phillips, D. C., & Sarma, V. R. (1965).Structure of hen egg-white lysozyme. A three-


dimensional Fourier synthesis at 2 Å resolution. Na-ture (London) 206, 757–761.

2. Blake, C. C. F., Fenn, R. H., Johnson, L. N., Koenig, D. F.,Mair, G. A., North, A. C. T., Oldham, J. W. H., Phillips,D. C., Poljak, R. J., Sarma, V. R. & Vernon, C. A., (2001).Historical perspective. How the structure of lysozymewas actually determined. In International Tables for

Crystallography (Rossmann M. G. & Arnold, E., eds),vol. F. Kluwer Academic Publishers, Dordrecht,pp. 744–772.

3. Pauling, L., & Corey, R. B. (1951). Configurations ofpolypeptide chains with favored orientations aroundsingle bonds: two new pleated sheets. Proc. Natl Acad.Sci. USA. 37, 729–740.

50 Years of Protein Structure Analysis

Documents

Transcript of 50 Years of Protein Structure Analysis