Proc. NatL Acad. Sci. USAVol. 79, pp. 5263-5267, September 1982Biophysics

A model for catabolite activator protein binding to supercoiled DNA(gene expression/protein-DNA interactions)

F. R. SALEMME*Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, Connecticut 06511

Communicated by David R. Davies, May 27, 1982

ABSTRACT Catabolite activator protein (CAP) is a dimericmolecule (Mr = 2 X 22,500) involved in transcription initiation ofseveral catabolite-sensitive genes of Escherichia coli The presentcommunication proposes a model for the interaction of CAP withDNA. The model is based upon known geometrical features of theCAP molecule [McKay, D. B. & Steitz, T. A. (1981) Nature (Lon-don) 290, 744-749], which allow interaction between dyad-relateda-helices of the dimeric protein and major grooves in adjacentlyaligned sections of right-handed B-DNA. These geometrical fea-tures suggest that in vivo CAP binding to closed-circular DNA in-volves CAP bridging adjacent loops of a DNA solenoidal coil. Thisinteraction pattern is shown to be consistent with the geometricaland stoichiometric properties of nonspecifically bound CAP com-plexes observed by Chang et aL [Chang, J. J., Dubochet, J., Bau-dras, A., Blazy, B. & Takahashi, M. (1981) J. MoL BioL 150,435-439]. CAP-induced coil formation is related to in vivo CAPpotentiation of RNA polymerase activity in underwound closed-circular DNA. Specifically, it is proposed that CAP binding to theright-interwound form of supercoiled DNA effects a local redis-tribution of DNA twist-strain energy, thus resulting in the for-mation of a left-handed solenoidal loop. The production of this lo-calized solenoidal loop, which reflects compensatory alterationsin DNA twist and writhe, may provide a conformationally uniquesite for RNA polymerase binding where the DNA is partially un-wound. The proposed interaction pattern is consistent with bothrecent DNA unwinding experiments and various nuclease protec-tion data. Moreover, features of the model suggest that the re-petitive and symmetric character of many promoter sequencesmay provide the structural basis for a switching mechanism op-erative in the differential control of gene transcription.

Considerable interest attaches to the structural nature of inter-actions between proteins and DNA. Recently, x-ray structuresoftwo DNA binding proteins involved in the regulation ofgeneexpression have been reported, the bacteriophage A cro re-pressor protein (1), which is involved in the lytic phase ofphagedevelopment (2-5), and the Escherichia coli catabolite activatorprotein (CAP) (6), which functions in the cyclic AMP-dependentexpression of several catabolite-sensitive gene operons (7-10).Although CAP and cro differ considerably in their overall ter-tiary folding, both molecules are active as dimers and incor-porate individual subunits (1) or functional domains (6) relatedby two-fold rotational symmetry. This feature suggested modelsfor the interactions of these proteins with DNA, which in bothcases involved the packing of symmetry-related a-helices intothe major grooves ofDNA (11). In the case ofcro repressor, themolecular dyad relates two antiparallel a-helices whose axes areabout 30 A apart and so situated with respect to the remainderof the molecule to allow intimate interaction between the he-lices and the major grooves ofright-handed B-DNA (1) (Fig. la).The CAP dimer also incorporates a symmetrical pair of anti-

parallel helices whose axes are about 30 A apart. However, de-spite recent evidence suggesting that the DNA binding domainsofCAP and cro may be evolutionarily related (12), the moleculardimers otherwise differ in the relative orientations oftheir dyad-related a-helices. This is schematically shown in Fig. 1 a andb, which illustrate the approximate mirror relationship betweenthe a-helix orientations in the CAP and cro molecules. Cor-respondingly, it was suggested (6) that CAP protein binds a left-handed form of B-DNA (Fig. lb). Although recent structuralstudies of left-handed forms ofZ-DNA (13, 14), as well as the-oretical studies of left-handed B-DNA conformations (15, 16),suggest that such alternative structures are energetically rea-sonable, unwinding experiments as yet give little evidence forleft-handed regions in naturally occurring DNA (17-19). Thepresent communication describes an alternative model for CAPDNA binding which preserves features in common with themodel previously described (6) but instead involves CAP in-teraction with right-handed B-DNA. The fundamental featureof this model is illustrated in Fig. ic. Here, two aligned right-handed DNA helices form a pattern of parallel major grooveswhich resembles that of a single left-handed helix of otherwisesimilar geometry. Consequently, a single CAP molecule canbind to this arrangement so that each CAP a-helix makes majorgroove interactions with sections of right-handed B-DNA. It isproposed that in vivo interactions between CAP and closed-cir-cular DNA induce the formation of solenoidal DNA supercoils(Fig. 2), thus resulting in the concomitant production of bindingsites similar to those shown in Fig. ic. The proposed bindingmode suggests several novel features of the CAP-DNA inter-action that are related to both the facilitation of RNA polymer-ase binding and subsequent gene transcription.

Nonspecific CAP-DNA binding as a structural precedentfor CAP-induced supercoil formation

A bridging CAP interaction of the sort illustrated in Fig. icnecessitates the adjacent alignment of two sections of DNAdouble helix. Evidence for the structural nature of this inter-action, as might be associated with CAP binding to closed-cir-cular DNA in vivo, is suggested by recent electron microscopystudies of nonspecific CAP-DNA complexes (20). Althoughsequence-specific CAP binding occurs only in the presenceof cyclic AMP, CAP exhibits nonspecific DNA binding with1/100th to 1/1,000th the affinity (21).

Current proposals for protein-DNA complex formation en-vision the participation of both sequence-specific and non-se-quence-specific interactions between the protein and DNA.Examples of the former include interactions between proteinside chains and accessible base-pair hydrogen bonding groups

Abbreviation: CAP, catabolite activator protein.* Present address: Dept. of Biochemistry, Univ. ofArizona, Tucson, AZ85721.

5263

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

Proc. Natl. Acad. Sci. USA 79 (1982)

g1K

I-

c

a b

FIG 1. (a) Proposed mode of cro repressor binding to right-handedB-DNA (1). Two dyad-related a-helices of the cro dimer (shown as cyl-inders) fit into successive major grooves in right-handed B-DNA. Al-though successive major grooves of the DNA (shaded) repeat every 34A along the direction of the DNAhelix axis, a-helices oriented to followthe pitch of the grooves pack optimally when they are parallel andhavetheir axes separated by about 30 A. (b) Arrangement of dyad-relatedhelices in the CAP molecule. These bear an approximate mirror re-

lationship relative to that observed in cro, and so appear to interactoptimally with a left-handed form of B-DNA (6). (c) The major groovesof two adjacent right-handed B-DNA helices can be readily aligned inparallel fashion to produce a pattern of grooves between helices whichresemble that of a left-handed helix. CAP can consequently bind tosuch an arrangement without necessitating the formation of left-handed DNA.

in either major or minor DNA grooves (22, 23), whereas thelatter may involve either ionic and hydrogen bonding interac-tions with groups of the deoxyribose backbone or generalizedhydrophobic packing effects between the protein and groupsin the DNA grooves (24). Available data on the relative mag-nitudes of sequence-specific vs. nonspecific binding constantssuggest that nonspecific interactions make major contributionsto the free energy ofcomplex formation (24) and, moreover, mayplay an important role in diffusional search processes ofproteinsbound to DNA (25). Consequently, non-specific CAP bindingin the absence of effector cyclic AMP might reasonably be ex-

pected to manifest interactions characteristic offunctional bind-ing, as consistent with a difference in specific vs. nonspecificDNA binding constant corresponding to the formation free en-

ergy of one to two additional hydrogen bonds.Electron microscope examination of DNA completely dec-

orated with CAP (20) shows long cylindrical rods 110 A indiameter, and optical diffraction patterns suggest the structureis organized as a long stack of helically dislocated disks about41 A thick. These data, taken together with the measured 4: 1length ratio of native to complexed DNA length, suggest thatthe DNA is organized as a tightly wound solenoidal supercoil.

FIG. 2. CAP binding to right-handed closed-circular DNA can oc-cur when the DNA forms a solenoidal supercoil. Although specific in-teractions between CAP and DNA could potentially occur with CAPbound to the inside of the solenoidal loop, observed properties of thenonspecific DNA complex suggest CAP binds to the outside of the coilas illustrated. Similarly, both structural considerations and unwind-ing experiments suggest the supercoil is left-handed.

One possible arrangement might involve a DNA supercoil witha pitch of 41 A decorated with bound CAP molecules. Alter-natively, a solenoidal structure can be envisioned in which the41-A disc repeat corresponds to two solenoidal DNA wraps (Fig.3). This necessitates that the DNA have an effective packingdiameter ofabout 20 A, which is somewhat less than the packingdiameter of 25 A observed for B-DNA in hydrated fiber dif-fraction studies (26), but closely approximates the value ex-pected for a dehydrated specimen (27). This latter arrangementis ofinterest in the current context, because model studies showthat it can be decorated with CAP molecules so that each dimera-helix makes extensive contacts with major grooves on adja-cent DNA supercoil loops. As detailed in Fig. 3, this arrange-ment can account for the observed geometric and stoichiometricproperties of the complex, and it approximates a stack of heli-cally dislocated discs that repeat with twice the axial repeat ofthe DNA (-2 x 20 A).The unusual feature of this structure is the very tightly

wrapped solenoidal conformation apparently attained by theDNA as a consequence ofCAP binding. Indeed, this is virtuallythe smallest radius supercoil structure possible for duplex DNAthat can accommodate bridging CAP molecules. While DNAconformational studies suggest the structural feasibility of thistight supercoil (28), estimates of the bend strain energy (29)

a b N

FIG. 3. (a) A model left-handed solenoidal coil (of right-handedDNA) decorated with nonspecifically boundCAP molecules. (b)A helixnet of the complex (i.e., the cylindrical surface unrolled to form a flatsheet), with a single turn of the DNA shaded. This model closely ap-proximates both the stoichiometric and geometrical properties of com-plex seen in the electron microscope. Specifically: (i) The observedlength ratio of complexed to uncomplexed DNA is 1: 4, necessitatingan extremely small supercoil radius of -16 A (supercoil axis to DNAhelix axis). This is significantly smaller than that observed for nu-cleosomes (-50 A) and is the smallest sterically allowable arrange-ment that can incorporate bridging interactions (-30 A) with dyad-related CAP a-helices. The loop spacing in alternative coil models (41A pitch), in contrast, appears too long to accommodate a bridging CAPinteraction in which a-helices fit major grooves. (ii) Making appro-priate allowance for the actual dimensions of the remainders of theCAP dimer (omitted in these figures for clarity), the model reasonablyapproximates the 11o-A diameter observed. (iii) All accessible majorgrooves accommodate a-helices of bound molecules, thus approximat-ing the observed stoichiometry of 1 CAP to 18 DNA base pairs. (iv)Filling all DNA major grooves with CAP a-helices necessitates thatthe cap molecules be organized as a right-handed double helix of mol-ecules (broken line in b), having an apparent axial repeat (-41 A) ofabout twice the dehydrated DNA helix diameter ("20 A). Stain pen-etration in regions having noncontiguous bridging CAP interactionscould cause this arrangement to appear as a helically dislocated stackof discs -41 A thick. (M) The arrangement reproduces the observed 3± 0.6 CAP molecules per 41-A repeat.

5264 Biophysics: Salemme

4",1..-<

L,--

Proc. Natl. Acad. Sci. USA 79 (1982) 5265

amount to some 2 kcal/mol of base pairs (1 keal = 4.184 kJ).Although this is a high value relative to that estimated for nu-cleosomes (30) and may reflect deviations from theory for highlybent coils, it is possible that multidentate protein-DNA inter-actions could yield energies of sufficient magnitude to bendDNA to such small radii. Such effects on DNA flexibility are,for example, manifest in the considerable decrease in DNApersistance length at high ionic strength (29). The importantpoint is that while CAP molecules could bridge solenoidal DNAof any radius consistent with appropriate alignment of majorgrooves on adjacent coils (i.e., at 30-A intervals), the observedcomplex corresponds to a coil of minimum radius. This impliesthat CAP binding induces the DNA conformational transitionand that the interaction energy of CAP binding compensatesfor the DNA deformation energy associated with supercoilformation.

Functional significance of CAP-induced supercoil formation

In the presence ofthe allosteric effector cyclic AMP, CAP bindsspecifically to recognition sequences near the promoters forseveral gene operons in E. coli. This specific CAP-DNA inter-action facilitates binding ofRNA polymerase at an adjacent site,subsequently initiating gene transcription (7-10). An observa-tion of particular relevance in the current context is that CAP-induced RNA polymerase binding is greatly potentiated in su-percoiled DNA (6, 21). As shown in Fig. 4, underwound closed-circular DNA has alternative superhelical conformations, aright-coiled interwound form (a) and a left-coiled solenoidalform (c). The energetic preference for the right-coiled inter-wound form is a consequence of the fact that this arrangementminimizes the local twist deformation in the supercoiled DNAassociated with closure ofan underwound circular DNA duplex(31, 32). However, the apparent binding behavior observed inthe nonspecific CAP-DNA complexes suggests that CAP bind-ing may induce the formation of a left-handed solenoidal coilin a localized region ofDNA with right-handed interwound con-formation (Figs. 2b and 4b). That is, binding of CAP to DNAeffects a redistribution of twist strain energy that results in analtered conformation of the DNA (see Fig. 5). The dependencyof CAP-induced solenoidal coil formation on the underwoundsupercoiled state of the DNA suggests that this conformationalchange is associated with compensatory local alterations in DNAwrithe and twist (31, 32). In this case ALk may remain constantwhen -AW = AT, in which Lk, W, and Tare, respectively, thelinkage; writhing, and twist values of the DNA. Such a com-pensatory change in local DNA conformation is indeed sug-gested by recent experiments by Kolb and Buc (21), which show

a b c

FIG. 4. The most energetically favorable conformation for an un-derwound closed circular DNA duplex is the right-interwound super-coiled form shown in a. However, this is in equilibrium with a left-handed solenoidal form shown in c. It is proposed that the binding ofCAP induces a local conformational transition between these forms,to produce a localized solenoidal arrangement (b) whose detailed struc-ture is schematically illustrated in Fig. 2. Note that this interconver-sion reflects a compensatory alteration in DNA writhe and twist, sothat the alternative conformations (a, b, and c) have constant linkagenumbers.

FIG. 5. Owing to the helical twist ofDNA, side chains of an a-helix(which is basically a cylindrical object) can interact with sequence-spe-cific features in the DNA major groove over region of four to seven basepairs. For example, side chain groups of the a-helix might form hy-drogen-bonded pairs spanning four residues, which over a limited re-gion could form specific complementary hydrogen-bonding interac-tions with exposed base pair edges in the major groove. An interestingfeature of this interaction pattern is-that it improves as theDNA eitherbends or becomes untwisted-i.e., as the major groove better approx-imates a cylindrical surface complementary to the a-helix surface. Thisoccurs owing both to better geometric correspondence between theDNA base pair rise and the groups of the a-helix and to more extendedinteraction between the a-helix and the DNA. Such interactions maytherefore result in both untwisting and coiling of the DNA duplex, orwhat is geometrically equivalent, a compensatory alteration in DNAwrithe and twist. This is a chiral effect-i.e., an extended interactionbetween the right-handed DNA helical groove and an a-helix requiresthe a-helix to assume a right-supercoiled conformation. Conversely,straightening the a-helix while preserving the major groove interac-tions requires the DNA to assume a left-supercoiled conformation.

that CAP binding to nicked DNA, followed by ligation, is as-sociated with only small changes in DNA linkage number.

Discussion

The present model envisions CAP binding to sequentially non-contiguous regions ofDNA. Previous models, in contrast, havesuggested that CAP binds to successive major grooves, ofeitherright-handed (33) or left-handed (6) DNA. Although the sub-sequent CAP structure determination appears to rule out amodel involving equivalent interactions between CAP mono-mers and contiguous major grooves in right-handed DNA (6),both models appear equally consistent with DNA "protection"experiments. Moreover, an important feature of both modelswas an implied specific interaction between equivalent CAPmonomer domains and corresponding dyad symmetric DNAsequences. These properties are considered further in the con-text of the present model.

Various DNA modification and digestion studies show thatCAP binding alters the DNA reactivity over a region spanning

Biophysics: Salemme

Proc. NatL Acad. Sci. USA 79 (1982)

1l/2-2 DNA helical turns (10, 34-38). These regions of modifiedreactivity presumably incorporate particular sequences thatconfer CAP binding specificity. As shown in Fig. 5, an opti-mized interaction between an a-helix and sequence-specificfeatures exposed in the DNA major grooves suggests a recog-nition sequence of some four to seven base pairs. Indeed, se-quence comparisons (33, 36, 38) of a number of CAP bindingsites in the lac (34, 38), ara (10), and gal (36) operons show acommon sequence of 5'-T-G-T-G-A-3'. Nevertheless, the oc-currence of this consensus sequence in inverted form, whereit would present an equivalent dyad-related site in an adjacentDNA major groove, is not a feature common to the known CAPbinding sites. While this may reflect some flexibility in natureof the CAP-DNA interaction as manifest in its nonspecific bind-ing properties (20, 21), it rules out the possibility of exactlyequivalent interactions between CAP monomers and featurespresented in successive DNA major grooves of the apparentCAP binding sites (33).

However, sequence and modification studies on both the lac(34, 38) and ara (10) operons show the presence of a second CAPbinding site, whereas inspection of regions surrounding CAPsites in these and other genes show the recurrence of the sameor similar consensus sequences as both direct and inverted re-peats (36, 37). The relevance of these observations to the pres-ent model is made clearer by Fig. 6. Fig. 6a corresponds to alinear stretch of duplex DNA, which for illustrative purposesis shown to incorporate two dyad-symmetric CAP binding sites(a situation best approximated in lac) (38). As shown in Fig. 6b-d, formation of a solenoidal coil can allow CAP bridging in-teractions with these sites in a number of different ways. Note,however, that the specific a-helix-major groove interactionsdiffer in these arrangements, b and d incorporating different,but symmetric helix-groove interactions, and c incorporatingasymmetric helix-groove interactions (i.e., the relative NH2-to COOH-terminal senses of the a-helices differ for the twogroove interactions). Thus, it may be that only arrangement b

1 2 3 4

a1 2

3 4

1 2

3 4

1 2

3 4

FIG. 6. (a) A linear segment of helical duplexDNA in which majorgrooves have been shaded to correspond to an idealized pair of dyad-symmetric CAP binding sites. (b-d) CAP molecules binding to alter-native supercoiled conformations. As illustrated by the relationshipof the sense of the a-helices (arrows) and the groove shading pattern,helix-groove interactions in b and d symmetrically differ, whereas cincorporates both. This would imply different interaction energies forthese arrangements. Although these may be functionally relevant,actual sequences (as described in the text) generally differ from theidealized arrangement and typically exhibit CAP consensus elementsthat allow the formation of energetically equivalent interactions (e.g.,when the sequences in the consensus grooves are palindromic) for dif-ferent groove registration patterns. Such alternative arrangementsresult in different accessibilities to sequences adjacent bound CAPmolecules and so may provide the basis for control mechanisms inwhich CAP interacts or competes with other DNA binding proteins.

or d is a functionally significant interaction allowing equivalenthelix-groove interactions. This is in no way inconsistent withobserved alterations in DNA susceptibility to chemical modi-fication, because CAP-DNA interactions of any one helix-groovesymmetry could account for the observed protection patternsin the short DNA pieces used in such studies.CAP binding to DNA in the manner presented in Fig. 6 b

or d suggests a particularly simple resolution of a controversyconcerning how CAP activates transcription by RNA polymer-ase. The basic question concerns whether or not CAP promotespolymerase binding and transcription by direct protein asso-ciation on the DNA or, alternatively, CAP binding induces somepropagated DNA conformational alteration that promotes poly-merase binding in the absence of direct interprotein contact(5; 33). The origin of this controversy lies in the observation thatthe apparent CAP binding sites in different catabolite-sensitiveoperons are situated at various sequential positions relative tothe initiation sites of gene transcription (10, 34-36). However,CAP-induced supercoil formation provides a means for bringingsequentially separated DNA regions into close physical prox-imity. The present model suggests that CAP-induced supercoilformation can allow the formation of equivalent and specificCAP-RNA polymerase interactions irrespective of the lengthof the intervening linear DNA sequences. Moreover, the so-lenoidal coil formed on CAP association constitutes a structur-ally unique site on the DNA where it is partially unwound (31),an established factor in the potentiation of binding and tran-scription by RNA polymerase (6, 39).

Although Fig. 6 depicts an arrangement of equivalent dyad-symmetric binding sites, the actual sequences (as describedabove) show variations in the' proximity of CAP binding sites.This suggests that CAP can interact alternatively with differentoperons that result in formation of solenoidal coils of differentradii. More interesting, however, is the possibility that CAP caninteract alternatively with a given operon. Here the observationof interest is that catabolite-sensitive operons frequently incor-porate CAP consensus or related sequences either as direct re-peats or imbedded in slightly longer palindromic sequences(33). The effect of either of these occurrences is to provide al-ternative modes of specific CAP-DNA association with a givenoperon. For example, ifeach consensus sequence in Fig. 6werepalindromic, the specific a-helix-groove interactions wouldbecome independent of helix sense, and helix-groove interac-tions shown for all arrangements in Fig. 6 would become equiv-alent. The point ofthis simplest-case example is to illustrate thatthese alternative groove registration patterns result in differingaccessibilities of adjacent DNA sequences, CAP consensussites, or both. This property is of interest because it is knownthat CAP acts in concert with several other regulatory proteins,whose apparent recognition sequences overlap CAP bindingsites and which are involved in the differential control of genetranscription (5, 10). In other words, a given operon may pro-vide specific binding sites allowing CAP to bind to a numberof alternative coiled DNA configurations that differ in their rel-ative major groove registration patterns. The selection of a par-ticular CAP binding pattern may therefore depend upon thepresence or absence of additional effector proteins binding toadjacent DNA sequences, as proposed, for example, in the bi-directional control of transcription in the ara operon (10). Con-versely, the observed variations in the arrangement ofCAP con-sensus sequences in different operons may reflect adaptationsof a common CAP functional role to different requirements as-sociated with the differential control ofthese genes' expression.These observations suggest that the repetitive and symmetricalsequence properties seen in many operators do not simply re-flect binding sites for dimeric proteins but instead form the basis

5266 Biophysics: Salemme

Proc. Natd Acad. Sci. USA 79 (1982) 5267

of a switching mechanism operative in the differential controlof gene transcription.

In summary, it is proposed that CAP binding in vivo resultsin a redistribution of twist-strain energy in underwound DNA.This is associated with a compensatory alteration of DNA twistand writhe and the corresponding production of a structurallyunique site for polymerase binding. In this sense CAP can beviewed as an allosteric effector of a conformational change de-pendent on the supercoiled state of the DNA (40). Conversely,repressor molecules such as cro, which are proposed to interactwith successive grooves in DNA (1), might be viewed as neg-ative allosteric effectors of the DNA conformational change as-sociated with gene expression, as previously suggested (6).The author thanks T. A. Steitz, J. Rosa, and D. Crothers, together

with other members of the Biophysical Structure Group at Yale Uni-versity for stimulating and critical discussions, and J. L. Mouning forhelp with the photography. F.R.S. is on sabbatical leave from theUniversity of Arizona at Tucson.

1. Anderson, W. F., Ohlendorf, D. H., Takeda, Y. & Matthews, B.W. (1981) Nature (London) 290, 754-757.

2. Herskowitz, I. & Hagen, D. A. (1980) Rev. Genet. 14, 399-445.3. Herskowitz, I. A. (1973) Rev. Genet 7, 289-323.4. Echols, H. A. (1972) Rev. Genet. 6, 157-190.5. Ptashne, M., Jeffrey, A., Johnson, A. D., Maurer, R., Meyer, B.

J., Pabo, C. O., Roberts, T. M. & Sauer, R. T. (1980) Cel 19,1-11.

6. McKay, D. B. & Steitz, T. A. (1981) Nature (London) 290, 744-749.

7. Zubay, G., Schwartz, D. & Beckwith, J. (1970) Proc. NatL Acad.Sci. USA 66, 104-110.

8. de Crombrugghe, B., Perlman, R. L., Varmus, H. E. & Pastan,I. (1969) J. BioL Chem-. 244, 5828-5835.

9. Epstein, W., Rothman-Denes, L. B. & Hesse, J. (1975) Proc.NatL Acad. Sci. USA 72, 2300-2304.

10. Ogden, S., Haggerty, D., Stoner, C. M., Kolodrubetz, D. &Schleif, R. (1980) Proc. NatL Acad. Sci. USA 77, 3346-3350.

11. Zubay, G. & Doty, P. (1959)J. MoL BioL 1, 1-20.12. Steitz, T. A., Ohlendorf, D. H., McKay, D. B., Anderson, W. F.

& Matthews, B. W. (1982) Proc. Nati Acad. Sci. USA 79, 3097-3100.

13. Wang, A. H.-J., Quigley, G. J., Kolpak, F. J., Crawford, J. L.,Van Boom, J. H., Van der Marel, G. & Rich, A. (1979) Nature(London) 282, 680-686.

14. Drew, H., Takano, T., Tanaka, S., Itakura, K. & Dickerson, R.E. (1980) Nature (London) 286, 567-573.

15. Gupta, G., Bansal, M. & Sasisekharan, V. (1980) Proc. Nati Acad.Sci. USA 77, 6486-6490.

16. DeSantis, P., Morosetti, S. & Palleschi, A. (1981) Biopolymers20, 1727-1739.

17. Wang, J. C. (1979) Proc. Natl Acad. Sci. USA 76, 200-203.18. Peck, L. J. & Wang, J. C. (1981) Nature (London) 292, 375-378.19. Rhodes, D. & Klug, A. (1981) Nature (London) 292, 378-380.20. Chang, J. J., Dubochet, J., Baudras, A., Blazy, B. & Takahashi,

M. (1981) J. Mol Biol 150, 435-439.21. Kolb, A. & Buc, H. (1982) Nucleic Acids Res. 10, 473-485.22. Seeman, N. C., Rosenberg, J. M. & Rich, A. (1976) Proc. Natl

Acad. Sci. USA 73, 804-808.23. von Hippel, P. H. & McGhee, J. D. (1972) Annu. Rev. Biochem.

41, 231-300.24. von Hippel, P. H. (1979) in Biological Regulation and Develop-

ment, ed. Goldberger, R. F. (Plenum, New York), Vol. 1, pp.*279-347.

25. Winter, R. B., Berg, 0. G. & von Hippel, P. H. (1981) Biochem-istry 20, 6961-6977.

26. Langridge, R., Marvin, D. A., Seeds, W. E., Wilson, H. R.,Hooper, C. W., Wilkins, M. H. F. & Hamilton, L. D. (1960)J.Mole Biol. 2, 38-47.

27. Arnott, S. & Hukins, D. W. L. (1972) Biochem. Biophys. Res.Commun. 47, 1504-1509.

28. Olson, W. K. (1979) Biopolymers 18, 1235-1260.29. Record, M. T., Jr., Maxur, S. J., Melancon, P., Roe, J. H., Sha-

ner, S. L. & Unger, L. (1981) Annu. Rev. Biochem. 50, 997-1024.30. Bentley, G. A., Finch, J. T., Lewit-Bentley, A. & Roth, M.

(1981) in Structural Aspects of Recognition and Assembly in Bi-ological Macromolecules, ed. Balaban, M. (Balaban, Rehovot, Is-rael), Vol. 2, 683-688.

31. Bauer, W. R., Crick, F. H. C. & White, J. H. (1980) Sci. Am. 243(1), 118-133.

32. Calladine, C. R. (1980) Biopolymers 19, 1705-1713.33. O'Neill, M. C., Amass, K. & de Crombrugghe, B. (1981) Proc.

Natl Acad. Sci. USA 78, 2213-2217.34. Majors, J. (1979) Dissertation (Harvard University, Cambridge,

MA).35. Simpson, R. B. (1980) Nucleic Acids Res. 8, 759-766.36. Taniguchi, T., O'Neill, M. & de Crombrugghe, B. (1979) Proc.

Natl Acad. Sci. USA 76, 5090-5094.37. Musso, R. E., DiLauro, R., Adhya, S. & de Crombrugghe, B.

(1977) Cell 12, 847-854.38. Schmitz, A. (1981) Nucleic Acids Res. 9, 277-292.39. Dickson, R. C., Abelson, J., Barnes, W. M. & Reznikoff, W. S.

(1975) Science 187, 27-35.40. Gellert, M. (1981) Annu. Rev. Biochem. 50, 879-910.

Biophysics: Salemme