Proc. Nati. Acad. Sci. USAVol. 88, pp. 10148-10152, November 1991Biochemistry

The nucleosomal core histone octamer at 3.1 A resolution: Atripartite protein assembly and a left-handed superhelix

(nucleosome/chromatin/handshake motif/histone fold)

GINA ARENTS*, RUFUS W. BURLINGAME*t, BI-CHENG WANGt, WARNER E. LOVE§,AND EVANGELOS N. MOUDRIANAKIS*¶l*Department of Biology and §Thomas C. Jenkins Department of Biophysics, The Johns Hopkins University, Baltimore, MD 21218; $Departments ofCrystallography and Biological Sciences, The University of Pittsburgh, Pittsburgh, PA 15260; and IDepartment of Biology, University of Athens,Athens, Greece

Communicated by Christian B. Anfinsen, August 30, 1991

ABSTRACT The structure of the octameric histone core ofthe nucleosome has been determined by x-ray crystaflography toa resolution of 3.1 A. The histone octamer is a tripartite assemblyin which a centrally located (H3-H4)2 tetramer is flanked by twoH2A-H2B dimers. It has a complex outer surface; depending onthe perspective, the structure appears as a wedge or as a flat disk.The disk represents the planar projection of a left-handed pro-teinaceous superhelix with -28 A pitch. The diameter of theparticle is 65 A and the length is 60 A at its maximum and -10A at its Ium extension; these dimensions are in agreementwith those reported earlier by Klug et al. [Klug, A., Rhodes, D.,Smith, J., Finch, J. T. & Thomas, J. 0. (1980) Natwre (London)287, 509-516]. The folded histone chains are elongated ratherthan globular and are assembled in a characteristic "handshake"motif. The individual polypeptides share a common centralstructural element of the helix-oop-helix type, which we namethe histone fold.

In all eukaryotic cells, the nuclear DNA is highly compactedthrough its association with special proteins, the histones,which both neutralize its electrostatic character and providea scaffold for its path. This path must be compatible with theseveral states of organization the double helix experiences asit progresses from interphase chromatin to the metaphasechromosome and back. At the first level of compaction, thehistones and DNA are organized in repeating units (1) callednucleosomes (2). Within each core nucleosome are twocopies of each of the core histones H2A, H2B, H3, and H4in the form ofan octameric complex (3). The histone octamerhas been shown by a variety of solution physical chemicalexperiments to be internally organized as a tripartite proteinassembly (4, 5).Over the past decade, the structure ofboth the nucleosome

and the histone octamer has been the subject of intensiveinvestigations. They have included small-angle neutron andsmall-angle x-ray diffraction (6, 7) of solutions of theseassemblies, as well as x-ray crystallography of their orderedstates crystallized with the aid of either organic solvents (8,9) or high salts (10, 11). Our earlier crystallographic analysisof the chicken erythrocyte histone octamer described theoctamer as having a "tripartite organization, that is, a central(H3-H4)2 tetramer flanked by two H2A-H2B dimers." Itsshape was described as an ellipsoid 110 A long and 65-70 Ain diameter (11). At that level of analysis, the internalorganization of the particle was in agreement with our earlierbiochemical work that demonstrated the tripartite nature ofthe histone octamer in solution (4) and could accommodatethe histone-DNA contacts suggested by the studies of Mirz-abekov et al. (12). However, the shape and size of our model

differed drastically from those derived from the crystallo-graphic studies of Finch et al. (8) with the nucleosome coreparticle, and from those of Klug et al. (13) on the histone coreof the nucleosome (hereafter called the MRC model). In thisreport, we present the results of our crystallographic rede-termination of the structure of the histone octamer. It willbecome apparent that the "shape and size" issue is nowresolved; the outer dimensions of the histone octamer areconsistent with the MRC model (length, 55 A; diameter, 70 A)(13) and not with those reported earlier by us (length, 110 A;diameter, 70 A) (11). The shape of the octamer is complexand, depending on the perspective, it can be described as awedge or a flat disk (13) as well as a "tripartite assembly witha central V-shaped (H3-H4)2 tetramer, flanked by two flat-tened balls, the H2A-H2B dimers" (11). The increased lengthwe reported earlier was the result of an apparent rotation ofthe electron density in the Fourier map about a "specialposition." This rotation, at the end, resulted in a much"expanded" tetramer and distally displaced dimers, thusgenerating an octamer with the erroneous length of 110 ratherthan 60 A. An extended and detailed account of the crystal-lographic analyses germane to this transformation will bepublished elsewhere (B.-C.W., J. Rose, G.A., and E.N.M.,unpublished data). The results of our redetermination of thestructure of the histone octamer to a resolution of 3.1 A andan R value of 25.5% are presented here at two levels.** First,we present solid renderings of the shapes of the histoneoctamer to facilitate direct comparison with the earlier lower-resolution studies (8, 11, 13). Second, we outline salientrecently found features of the folding of the four polypeptidesas well as their organization within the H2A-H2B dimer and(H3-H4)2 tetramer subunits of the histone octamer. A de-tailed analysis will be presented in a forthcoming article.

MATERIALS AND METHODSStructure Determination. The chicken erythrocyte histone

octamer forms large crystals in space group P3221 thatdiffract to better than 3.0-A resolution (10). Only one-half ofthe octamer-i.e., H2A-H2B-H3-H4-is present in the asym-metric unit. This requires the molecular and crystallographictwofold axes to coincide. A heavy atom derivative of thehistone octamer crystals with tetrakis-(acetoxymercu-ri)methane was found that gave only one peak in the Patter-son map (11) at x = 0.322, y = 0.329, z = 0.992 (equivalent

Abbreviation: ISIR, iterated single isomorphous replacement.tPresent address: Scripps Clinic and Research Foundation, La Jolla,CA 92037."To whom reprint requests should be addressed at: Department ofBiology, The Johns Hopkins University, Baltimore, MD 21218.**The a-carbon coordinates have been deposited in the Protein DataBank, Chemistry Department, Brookhaven National Laboratory,Upton, NY 11973 (reference 1HIO).

10148

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

7, 2

021

Proc. Natl. Acad. Sci. USA 88 (1991) 10149

to z = -0.008), and the occupancy of this site was <50osince the two -SH groups of the octamer are very close toeach other on either side of the dyad. Thus, only one metalatom can be bonded to one but not to both of the cysteines.Phases calculated by the iterated single isomorphous replace-ment (ISIR) procedure of Wang (15) resulted in the electrondensity map shown earlier (11).For reasons to be detailed elsewhere (B.-C.W., J. Rose,

G.A., and E.N.M., unpublished data), the data were reex-amined and the heavy atom parameters were reevaluated.This process resulted in a slight displacement of the heavymetal position to a new site with x = 0.342, y = 0.344, z =0.007, and an improvement of the R(culliS) value for this sitefrom 51.3% to 48.0%o. Further examination of our earlierresults revealed that the minimum in R value calculated forthe old site was actually a local minimum. Comparison of theheavy metal difference Patterson (FPH - Fp)2 with theoreticalPatterson syntheses based on the old and new sites confirmedthe correctness ofthe new site; therefore, the old heavy metalcoordinates were abandoned. The ISIR procedure was usedonce again to calculate new phases for the original data byusing the coordinates of the new heavy atom site, and a newelectron density map was calculated.By comparing the two Fourier maps, it is clear that the

molecular boundary found in the old map is larger than thatfound in the new map. The old boundary contains the samefundamental domains of electron density present in the newmap, as well as partial repeats of one of those domains. Inaddition, there exists a 1200 rotational displacement of themajor domains between the old and new maps; it was thisrotation about the special position that generated the falseimage in the old map and this resulted in an elongated shapefor the histone octamer (B.-C.W., J. Rose, G.A., andE.N.M., unpublished). We constructed our molecular modelwithin the new map for the following reasons. In this appli-cation of the ISIR procedure, we used modified heavy atomcoordinates that had refined to a better R value and thephases generated thereby had a better figure of merit (new(m) = 0.75; former (m) = 0.65); consequently, this map showsimproved connectivities between major secondary structureelements as well as improved side-chain densities (Fig. 1).

Finally, the asymmetric unit of the new map contains noredundant electron density-i.e., there is no partial repeat ofthe dimer density near the tetramer.

Initial Model. The interpretation of the map was facilitatedby the fact that the asymmetric unit contains half of the(H3-H4)2 tetramer and one H2A-H2B dimer and a singlecysteine as residue 110 ofH3. A simple inspection ofthe maprevealed one contiguous protein domain containing the mer-cury of tetrakis-(acetoxymercuri)methane (11) and this wasassigned to the H3-H4 half-tetramer. The other separate butclosely apposed domain was assigned to the H2A-H2B dimer.It is interesting to note that the density of the dimer domainin the new map is nearly identical to the density for theanalogous dimer domain in the previous map. The overallquality of the present map and clear connectivities betweensecondary structure elements in it allowed for a straightfor-ward tracing of the four chains.We have built a model for the histone octamer by using the

program FRODO (16) on an Evans and Sutherland PS-390 andMicroVAX II minicomputer. Our initial model containedside-chain atoms for approximately two-thirds of the H2A-H2B residues, while all others were built in as alanines. Asingle round of refinement by simulated annealing using theprogram XPLOR (17) reduced the crystallographic R valuefrom 48% to 29% for the 10- to 3.5-A data set. An improvedmap was calculated by using the SIGMAA program of Read(18) to combine the model phases with the ISIR phases.Another round ofrebuilding, refitting, and refining using bothXPLOR and the restrained least-squares refinement programPROFFT (19-21) was initiated by using the 10- to 3.1-A dataset-i.e., 13,542 reflections with F. > 2 X o(Fo). Theresulting model (no water molecules included) has goodoverall stereochemistry (rms deviations from ideal bondlengths, 0.016 A; rms deviations from ideal bond anglelengths, 0.039 A), and 4-q values that are reasonable for apartially refined model (22)-i.e., <5% fall in energeticallyunfavorable regions. The R value for 2644 nonhydrogenatoms with individual thermal factors is 25.5% (thermalvariabilities between main chain atoms weighted to 1.5 A2).The difference Fourier map suggests that significant improve-ments to the model can be made in two areas: (i) by refittingthe extended chain regions between major helices, and (ii) byextending the model building into the areas of weak electrondensity distal to the current carboxyl end of H2A and aminoend of H2B.

RESULTSSecondary Structure. Our current model accounts for

=70% ofthe histone octamer, and ofthese residues =65% arefound in helices (Fig. 2). This is slightly in excess of the 62%assigned to structured residues by the earlier studies ofBradbury and associates (23-25). Within the interpreted part

H2A F - -00 DMOO _ o Omo0OO

H2B --------Wom oooo

H3 - - - - - - 00000

H4 F- - - g M o--m

FIG. 1. Stereo photograph demonstrating the quality of the ISIR

map. The long helix of H2A is shown, with some density below itattributable to H2B.

1 20 40 60 80 100 120 140Residue Number

FIG. 2. Secondary structure of the four core histone chains.a-Helices are represented by coils, extended chains are representedby solid lines, and unseen residues are represented by dotted lines.Boldface segments of each chain participate in the common histonefold.

Biochemistry: Arents et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

7, 2

021

Proc. Natl. Acad. Sci. USA 88 (1991)

of the map the density for most side chains is excellent (Fig.1), with the expected exception of truncated density for theside chains of 30 polar residues. Since the poorly imaged andmissing portions of the electron density are not expected tocontribute additional secondary structure, we calculate that-45% of the total octamer residues are found in helices, inagreement with biochemical studies (refs. 26 and 27 andreferences therein). Of the remaining 35% of the residues inthe model, 10%o are in P-strands, mostly found in intermo-lecular contacts. The remaining 25% of the model is in arandom-coil configuration.

Description of the Octamer StrUCture. The histone octamer is atripartite protein assembly consisting of two H2A-H2B dimersflanking one centrally located (H3-H4)2 tetramer. Depending on

I

FIG. 4. Left-handed protein superhelix. The first dimer is at theupper left and sits behind the tetramer, which curves forward at thebottom left, crosses the center, and curves back touching the seconddimer at the upper right. The dimers are dark blue; the firsthalf-tetramer is gray; the second half-tetramer is white. In this view,the histone octamer can be perceived as tetrapartite. Black linerepresents the path of the left-handed protein superhelix.

FIG. 3. Three orthogonal views of the histone octamer. (a) Viewshowing the tripartite nature of the histone octamer, looking downthe molecular twofold axis with the superhelical axis running hori-zontally from left to right. We refer to this as the front view. (b)Protein wedge as it appears by looking down at a plane containing thetwofold and the superhelical axes, with the twofold axis running fromtop to bottom. The apex of the wedge is formed exclusively by thetetramer, while the H2A-H2B dimers form the lobes ofthe wedge. (c)View showing the histone octamer as a disk, looking down into thesuperhelical axis with the twofold axis horizontal. Protrusions fromthe curved surface are due to the termini of H2A, H2B, and H3.Surfaces were calculated from a-carbon positions at twice the vander Waals radius. The H2A-H2B dimers are dark blue and thetetramer is white.

the perspective of viewing this assembly, three distinct shapescan be perceived and the overall mass distribution can bedescribed as tripartite or tetrapartite. The tripartite image is mostclearly seen when looking straight into the dyad axis as it entersthe tetramer apex; we refer to this as the front view (Fig. 3a). Thecentrally located larger mass, which from this perspective ap-pears laterally biconcave and resembles a left-handed propeller,is the (H3-H4)2 tetramer. The two smaller masses, one on eachside ofthe tetramer, are the H2A-H2B dimers. The surface oftheoctamer is traversed by several grooves and ridges, which appearto follow the path of a left-handed superhelix. The axis of thissuperhelix is perpendicular to the twofold axis and runs horizon-tally from left to right in Fig. 3a. A perspective orthogonal to theearlier view and with the molecular dyad running from top tobottom reveals the octamer as a tripartite wedge (Fig. 3b). Whenviewed down the superhelical axis (Fig. 3c), the octamer resem-bles a disk =65 A in diameter. A closer examination of this viewreveals that the disk really represents the planar projection of atetrapartite, left-handed, proteinaceous molecular superhelix(Fig. 4) formed by the spiraling ofthe fourdomains (H2A-H2B)l/

3 4

FIG. 5. Stereo pair of the H2A-H2B and H3-H4 dimer domainswithin the histone octamer, viewed approximately down the super-helical axis so as to optimize visualization of the four chains and thehistone fold. For reasons of clarity, only one of each histone pair isshown. The amino end of each chain in the model is marked by anarrow.

10150 Biochemistry: Arents et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

7, 2

021

Proc. Natl. Acad. Sci. USA 88 (1991) 10151

(H3-H4)1/(H3-H4)2/(H2A-H2B)2 along a central axis. The orderof these four histone domains on the protein superhelix is inagreement with that proposed by Klug et al. (13). Viewed thisway, the protein masses line the inside of an imaginary cylinderwith a diameter of 65 A and occupy one and two-thirds turnsabout its axis. This results in a protein trapezoid or wedge witha thin side (-10 A) at the tip and a thick side (-60 A) at the lobesof the wedge.Tetramer Structure. The tetramer contains two elements of

identical volume, each consisting of one H3 and one H4molecule. Each H3-H4 dimer (or half-tetramer) resembles acrescent-shaped sector, like a sector of an orange, but withone blunt end and one pointed end. The pointed ends of thetwo sectors establish strong contacts as they overlap at thetwofold axis and form the tetramer. Each sector is rotatedaway from the twofold axis by -15°; therefore, the entiretetramer resembles a shallow horseshoe that has been par-tially twisted open. As measured along the axis of themolecular superhelix, the twofold axis, and their mutualperpendicular, the entire tetramer measures about 46 x 52 x66 A. When viewed down the axis of the protein superhelix,the tetramer occupies =270° of the cylinder. A large centralcavity is bounded by the interior curvature of the tetramer onone face and by the dimers on the other faces. The openingsfrom the solvent into this cavity are roughly circular, with anapproximate diameter of 8 A. The current estimate for thevolume of the cavity is -5500 A3. At our present resolutionand level of refinement, we find traces of electron density inone portion of the cavity that is near the start of the orderedportion of H2B; therefore, we expect that in our final modelsome of the volume of the cavity may be filled by additionalresidues.Dimer Structure. The H2A-H2B dimer has some similarity

in shape and volume to the half-tetramer domain. Whenviewed in the coordinate system described above, the dimermeasures approximately 31 x 50 x 44 A. The outside of thedimer is not as smoothly curved as the half-tetramer; it canbe better described as a somewhat flattened and elongatedball rather than as a crescent-shaped sector. In the proteinsuperhelix of the octamer, each dimer occupies -160°.Within both the H3-H4 (half-tetramer) and the H2A-H2B

dimer domains, the pairwise association of the folded histonechains follow a characteristic "handshake" motif; that is,rather than assembling like the globular domains of the a and/3 chains of the hemoglobin dimer, which have small localcontacts, the histone chains, by clasping each other, developan extensive molecular contact interface (Fig. 5). Also,within both the H2A-H2B and the H3-H4 domains, theindividual polypeptide chains are folded in a somewhatsimilar manner, most noticeably in the central portion of eachchain. The common motif, which we call the histone fold,consists of a long central helix flanked on either side by a loopsegment and a shorter helix (Figs. 2 and 5). This structuralsimilarity suggests a common evolutionary origin for the fourcore histones, a finding supported by a mild primary se-quence homology among these four chains (ref. 28 andreferences therein). The separation of the amino- and car-boxyl-terminal regions by a long (24-28 residues) helix isreminiscent of the shape of the troponin family of structures,but the histone loops are shorter than those found in EFhands (cf. ref. 29) and do not contain the necessary aminoacid residues for binding calcium. The greatest differences inthe folding of the ordered portions of the histone chains arefound at their amino and carboxyl ends.

DISCUSSIONCorrelations. The overall size and shape of the 3.1-A-

resolution octamer structure we describe here is entirelyconsistent with the results of the 22-A-resolution study of

Finch et al. (8) and Klug et al. (13), the 16-A-resolutionneutron diffraction study of Bentley et al. (30), and the7-A-resolution x-ray diffraction study ofRichmond et al. (31).These reports, as well as our current work, show the proteinportion of the nucleosome to be a wedge-like structure witha persistently curving outer surface that resembles a helicalramp. Most of the mass in this protein assembly is concen-trated within four major segments, a feature also clearlynoted in the earlier work of Bentley et al. (30). We haveexplained how the tripartite and wedge views of the octamerare alternate perspectives of this assembly.The main area of difference between our interpretation of

the present histone octamer map and the model presented byRichmond et al. (31) is a result of the difference in resolutionof the two studies and concerns the location of the individualhistone molecules within the octamer. At the original 22- andsubsequent 7-A-resolution structures of the MRC group, itwas not possible to assign side-chain densities correspondingto the histone sequences and thus directly identify theindividual histone chains. To assign locations for the histonemolecules within the nucleosome structure, Klug et al. (13)used information from the histone-DNA cross-linking stud-ies of Mirzabekov et al. (cf. ref. 12) and from severalhistone-histone cross-linking studies (cf. ref. 13). To this theyapplied the straightforward minimum assumption that thecentral portion of each histone chain (excluding the highlycharged chain termini), because of its small size, consisted ofa single (roughly globular) domain. At 3.1-A resolution, wecan now trace the four unique polypeptide backbones as wellas identify the side-chain densities and place the histoneamino acid sequences. None of the four core histones iscompacted into a single globular domain. Instead, each chainis folded in a rather elongated fashion; upon assembly intotheir physiological subunits, the domains of the folded poly-peptides interdigitate extensively rather than each chainoccupying a unique and contiguous segment on the surface ofthe octamer. This arrangement generates the potential forseveral noncontiguous contacts between each of the fourpolypeptides and the DNA helix as it winds its path aroundthe octamer.

In addition, certain features of our model correlate wellwith several physical and biochemical studies. Martinson andcoworkers (32) showed that a cross-link between Pro-26 ofH2A and Tyr-40 of H2B could be induced by UV irradiationof dimers. We find that these two residues participate in oneof the numerous contacts between H2A and H2B. Martinsonet al. (33) have also found that three cross-links can be formedin intact octamers between the last 18 residues ofH4 and thecarboxyl-terminal half of H2B. The octamer model containstwo different types ofcontacts for H4 and H2B, both ofwhichinclude the residues identified above; the carboxyl terminusofH4 is located near the dimer-dimer interface, and, accord-ingly, the carboxyl-terminal one-third of H4 is able to formcontacts with both H2B molecules.The percentage and location ofordered protein found in the

new map also correlate well with the findings ofBradbury andcoworkers (23-25), who reported that residues 25-95 ofH2A,37-114 of H2B, 42-110 of H3, and 33-102 of H4 are ordered.Our model includes these residues and extends the structuredportion of the octamer by 66 more residues (Fig. 2).

Conclusions. The octameric histone core ofthe nucleosomeis a tripartite protein assembly that, depending on the per-spective of its viewing, appears more or less as a flat disk ora wedge. As noted by Baldwin et al. (6), Finch et al. (8), andKlug et al. (13), the histone octamer serves as a protein"spool" around which 140 base pairs of right-handed DNAis wrapped in the form of a left-handed DNA superhelix(DNA-SH) (inner diameter, -70 A; length, -55 A). In thepresent study, we have also resolved this core protein spoolas a left-handed protein superhelix (Pr-SH) (outer diameter,

Biochemistry: Arents et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

7, 2

021

Proc. Natl. Acad. Sci. USA 88 (1991)

-65 A; extension along the superhelical axis, -60 A). Theouter surface of this fairly evenly curved Pr-SH has regularlyspaced ridges and valleys that define a strong left-handedpath of -28-A pitch, very suggestive of the path of theDNA-SH in the nucleosome. The left-handed superhelicalspool of protein is formed by the ordered spiraling assemblyof one H2A-H2B dimer to one side (left, with the front asdefined earlier in Fig. 3a) of one (H3-H4)2 tetramer and asecond to the other side (right) of the tetramer. It should benoted that the area of the dimer-tetramer contact interface ismuch more extensive than the interface between the twoH3-H4 half-tetramers. However, the inter-dimer-tetramerinterface is more open and potentially accessible to solventthan the intra-tetramer interface. Equilibrium binding andcalorimetric studies have already established that this exten-sive dimer-tetramer interface is by far the preferred surfacefor the first octamer disassembly step (4, 5) and thus isenergetically less stable than the smaller intra-tetramer in-terface. The intersection of each of the two dimer-tetramerinterfaces with the surface of the octamer define an "en-trance perimeter" of potential solvent channels leading intothe inside of the octamer. This is consistent with our earlierproposal of these interfaces as sites for the regulation ofchromatin compaction-decompaction processes (4).The overall structural features of the histone octamer lead

to interesting suggestions about possible selective pressuresin the evolution of deoxynucleoproteins. We have proposedearlier (34) that the process of chromatin assembly can beviewed as a two-step event; first, local histone-DNA inter-actions result in the local dehydration of the double helix,which in turn leads to the development of a local left-handedcurvature or bending of the DNA axis. Progressive andsequential addition of more histone assemblies along thelength of the double helix will compound this initially localcurvature and extend it to a contiguous curvature, whicheventually gives rise to a long-range left-handed DNA super-helix. This nucleoprotein superhelix is further stabilized andsimultaneously punctuated as a "string of beads" by meansof the short-range histone-histone interactions within thecore octamer. Thus, the histone masses, by virtue of theirassociation with this DNA superhelix, are expected to dis-tribute themselves in a protein superhelix of the same sense.In our earlier model studies of the compaction of pure DNAin the absence of any protein, we observed that reduction ofthe water activity by simple chemical means led to theformation of left-handed DNA supercoils with dimensionscompatible with those found in chromatin (35). We havesuggested that the histones must have been selected throughevolution to fit within those dimensions characteristic to theDNA supercoil and to facilitate DNA bending and supercoilformation by reducing the local water activity upon theirassociation with the double helix. Reciprocally, histone-histone interactions would have been optimized for thedegree of specificity and favorable energetics to harmonizewith the DNA superhelix. We propose that the selectivepressure of such interactions is responsible for the evolu-tionary stability of the individual subunits of the histoneoctamer and for the strength and stringency of the next levelof interactions of these subunits to form the histone octamer.In other words, the selective pressure in histone evolutionmust derive from the extensive interfacial contacts betweenthe histones in the octamer on one hand and the outer surfaceof the histone superhelix with the inside of the DNA super-helix on the other.

We are grateful to Drs. Mario Amzel and Wayne Hendrickson fortheir critical reading of the manuscript and many helpful suggestions.We thank Dr. Ernesto Freire for allowing us access to the IRISgraphics station and use of the MIDAS (14) software in the Biocalo-rimetry Center and Dr. John Rose for computational assistance. We

also thank Dr. David Schmickel for providing some of the crystalsused in this study and Dr. Andreas Baxevanis for numerous criticaldiscussions, his help with the illustrations, and his unrestraineddevotion to this project. We also acknowledge the support of theSupercomputing Laboratory of the National Cancer Institute (Fred-erick, MD). This work was supported by a grant from the NationalInstitutes of Health (GM-33495 to E.N.M.).

1. Kornberg, R. D. (1974) Science 184, 868-871.2. Oudet, P., Gross-Bellard, M. & Chambon, P. (1975) Cell 4,

281-300.3. Thomas, J. 0. & Kornberg, R. D. (1975) Proc. Natl. Acad. Sci.

USA 72, 2626-2630.4. Eickbush, T. H. & Moudrianakis, E. N. (1978) Biochemistry

17, 4955-4964.5. Benedict, R. C., Moudrianakis, E. N. & Ackers, G. K. (1984)

Biochemistry 23, 1214-1218.6. Baldwin, J. P., Boseley, P. G., Bradbury, E. M. & Ibel, K.

(1975) Nature (London) 253, 245-249.7. Perez-Grau, L., Bordas, J. & Koch, M. H. J. (1984) Nucleic

Acids Res. 12, 2987-2996.8. Finch, J. T., Lutter, L. C., Rhodes, D., Brown, R. S., Rush-

ton, B., Levitt, M. & Klug, A. (1977) Nature (London) 269,29-36.

9. Finch, J. T., Brown, R. S., Rhodes, D., Richmond, T., Rush-ton, B., Lutter, L. C. & Klug, A. (1981) J. Mol. Biol. 145,757-769.

10. Burlingame, R. W., Love, W. E. & Moudrianakis, E. N. (1984)Science 223, 413-414.

11. Burlingame, R. W., Love, W. E., Wang, B. C., Hamlin, R.,Xuong, N.-H. & Moudrianakis, E. N. (1985) Science 228,546-553.

12. Mirzabekov, A. D., Shick, V. V., Belyavsky, A. V. &Bavykin, S. G. (1978) Proc. Natl. Acad. Sci. USA 75, 4184-4188.

13. Klug, A., Rhodes, D., Smith, J., Finch, J. T. & Thomas, J. 0.(1980) Nature (London) 287, 509-516.

14. Ferrin, T. E., Huang, C. C., Jarvis, L. E. & Langridge, R.(1988) J. Mol. Graphics 6, 13-21.

15. Wang, B. C. (1984) Acta Crystallogr. A 40, C12.16. Jones, T. A. (1978) J. Appl. Crystallogr. 11, 268-272.17. Brunger, A. T., Kuryan, J. & Karplus, M. (1987) Science 235,

458-460.18. Read, R. J. (1986) Acta Crystallogr. A 42, 140-149.19. Hendrickson, W. A. (1985) Methods Enzymol. 115, 252-270.20. Finzel, B. (1987) J. Appl. Crystallogr. 20, 53-55.21. Sheriff, S. (1987) J. Appl. Crystallogr. 20, 55-57.22. Ramachandran, G. N., Ramakrishnan, C. & Sasisehkaran, V.

(1963) J. Mol. Biol. 7, 95-99.23. Bradbury, E. M. & Rattle, H. W. E. (1972) Eur. J. Biochem.

27, 270-281.24. Bradbury, E. M., Cary, P. D., Crane-Robinson, C. & Rattle,

H. W. E. (1973) Ann. N. Y. Acad. Sci. 222, 266-288.25. Moss, T., Cary, P. D., Abercrombie, B. D., Crane-Robinson,

C. & Bradbury, E. M. (1976) Eur. J. Biochem. 71, 337-350.26. Godfrey, J. E., Baxevanis, A. D. & Moudrianakis, E. N.

(1990) Biochemistry 29, 965-972.27. Beaudette, N. V., Fulmer, A. W., Okabayashi, H. & Fasman,

G. D. (1981) Biochemistry 20, 6526-6535.28. Brown, A. P. (1983) J. Theor. Biol. 104, 401-416.29. Strynadka, N. C. J. & James, M. N. G. (1989) Annu. Rev.

Biochem. 58, 951-998.30. Bentley, G. A., Lewit-Bentley, A., Finch, J. T., Podjarny,

A. D. & Roth, M. (1984) J. Mol. Biol. 176, 55-75.31. Richmond, T. J., Finch, J. T., Rushton, B., Rhodes, D. &

Klug, A. (1984) Nature (London) 311, 532-537.32. Callaway, J. E., DeLange, R. J. & Martinson, H. G. (1985)

Biochemistry 24, 2686-2692.33. Martinson, H. G., True, R., Lau, C. K. & Mehrabian, M.

(1979) Biochemistry 18, 1075-1082.34. Moudrianakis, E. N., Anderson, P. L., Eickbush, T. H.,

Longfellow, D. E., Pantazis, P. & Rubin, R. L. (1977) in TheMolecular Biology of the Mammalian Genetic Apparatus, ed.Ts'o, P. 0. P. (Elsevier North-Holland, Amsterdam), Vol. 1,pp. 301-321.

35. Eickbush, T. H. & Moudrianakis, E. N. (1978) Cell 13, 295-306.

10152 Biochemistry: Arents et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 2

7, 2

021