The phage 434 CroOR1 complex at 2.5 Å resolution

J. Mol. Hid. (1991) 219, 321-334

The Phage 434 Cro/O,l

Complex at 25 A Resolution

A. Mondragh2~t

and S. C. Harrison1~2

1 Howard Hughes Medical

Institute

2 Department of Biochemistry and Molecular Biology Harvard University, 7 Divinity Avenue Cambridge, MA 02138, U.S.A. (Received 13 July

1990; accepted 9 .Jan.uary 1991)

The crystal structure of phage 434 Cro protein in complex with a 20 base-pair D1VA fragment has been determined to 2.5 A resolution. The DNA fragment contains the sequence of the OR1 operator site. The structure shows a bent conformation for the DNA, straight,er at the center and more bent at the ends. The central base-pairs adopt conformations with significant deviations from coplanarity. The two molecules interact extensively along their common interface, both through hydrogen bonds and van der Waals interactions. The for operator binding and recognition is discussed. significance of these interactions

Keywords: DNA recognition;

DNA conformation; X-ray crystallography

1. Introduction The repressor and Cro proteins of phage 434 bind to a set of six related operator sites in the phage genome (denoted Oal-Oa3, O,l-0,3). Affinities of t.he two proteins for these six sites differ in such a way as to create an efficient regulatory switch 1986). Measurements of binding to (Ptashne, natural and synthetic operator sequences in vitro have established the essential outlines of differential recognition (Wharton et al., 1984; Koudelka et al., 1987, 1988). In brief, (1) a 434 operator site is a 14 base-pair sequence with the 4 base-pair element 5’-ACAA at either end; (2) the central six base-pairs of an operator site modulate its affinity for both repressor and Cro: (3) Cro differs from repressor in binding with nearly equal athnity to sites with 5’-ACAG and 5’-ACAA. Structural st,udies of the two proteins, alone and in complex with synthetic operator DNA, have begun to establish the mechanism by which these relative affinities are established. The 434 Cro and repressor DNA-binding domains are very similar, both in amino acid sequence and in threedimensional fold (Mondragon et al., 1989aJ). Both contain the now familiar helix-turn-helix motif. Two structures have been determined for complexes of the DSA-binding domain of 434 repressor, denoted Rl-69, with synthetic operators: a moderateresolution structure with a 14 base-pair fragment

t Present address: Department of Riochemistry. Molecular Biology and Cell Biology, Northwestern University. 2153 Sheridan Road. Evanston. TI, 60201. 1T.S.A.

helix-turn-helix;

repressor;

having a symmetric sequence related to 0,s (“14-mer”: Anderson et al., 1987) and a highresolution structure with a 20 base-pair fragment cont’aining the Oal sequence (Aggarwal et a,l., 1988). Both structures show that, two RI-69 monomers bind nearly symmetrically to an operator. Each tnonomer interacts with a half-site, and contacts between monomers create a significant dimer interface. A bound Rl-69 dimer fixes an operator in a bent conformation, overwound at its center and underwound near its ends. A significant narrowing of the minor groove occurs near the center. Contacts in the major groove between side-chains of the helix-turn-helix and groups on base-pairs 1 through 4 determine recognition of the 5’-ACAA element. Afhnity for different operators appears t’o depend on the ease with which the central nucleotIides can adopt. the required DNA conformat.ion. since there is no direct contact to the central four base-pairs. Aggarwal et (~2. (1988) have advanced specific proposals concerning the significance for different’ial affinity- of the non-Watson-Crick hydrogen bonds that can be made by non-coplanar base-pairs. A structure, determined at. moderate resolution for the complex of 434 Cro with the symmetric lbmer (Wolberger et al., 1988) showed a conformat’ion for the operator significantly different from that found in the complex with Rl-69. In particular, the 14-mer appeared straighter and more uniformly overwound when bound to Cro than when bound to Rl-69, and narrowing of the minor groove near the operator center was less pronounced than in the Rl-69 complex. The most obvious effect of the conformational differences on protein/DNA contacts was observed in the region near base-pair

322

.-I.

iMWLdnl.yci7L

0,

4, where the DNA backbone was much closer to the helix 2/helix 3 turn in Cro than to the turn in Rl-69. On the basis of these observations, Wolberger et al. (1988) suggested that the difference in imposed DNA conformations is critical to the different binding site “preferences” exhibited by the two proteins. To test this proposal, by examining the precise conformational differences between DNA operators in the two complexes, requires a higher-resolution structure than that afforded by the Cro/lkmer crystals. We report here the analysis at, 2.5 A resolution (1 A = 0.1 nm) of a complex between 434 Cro and the same (&-containing 20 base-pair fragment previously studied in complex with Rl-69. The structure shows that the conformation of the operator bound to Cro is indeed significantly different from its conformation when bound to Rl-69, but less extensively so than suggested earlier. The central 14 base-pairs in the 20 base-pair fragment closely resemble the 14-mer in complex with Cro (Wolberger et al., 1988), but the ends of the fragment curl around the Cro dimer t’o generate an overall bend very similar to that seen in the Rl-69 complex. Previous conclusions about differential bending by the two proteins (Koudelka et al., 1988; Wolberger et al., 1988) must therefore be revised. The high-resolution structure also shows how 434 (:ro establishes the observed operator conformation, and suggests that a number of differences between repressor and Cro, many of them not in helix 3, are important for differential recognition.

2. Experimental Methods (a) C’rystals

and data collection

Overproduction and purification of 434 Cro protein for crystallization have been described (Wolberger, 1987). DNA was purchased from PL-Biochemicals. The sequence is shown in Fig. 1. It is identical to that used in studies of a complex with Rl-69 (Aggarwal et al.. 1988). The central 14 base-pairs correspond precisely to 0,l. The oligonucleotide has unpaired bases at the 5’ end of each strand, so that base-pairs can form when adjacent fragments stack in a crystal (Jordan et al.. 1985). Crystals were

grown from 120j0 (w/v) polyethylene glpcol 3500. I00 ml~ Mes (pH 6.2), 160 mM-ka(I. 120 mM-Mg(‘l. 2 rnbb spermine using 2 mM-purified protein and I tnnl-DNA. The crystals grew at 4°C after a few days. Typical crystals are 500 pm x 250 pm x 100 pm. The space group is P2, with cell constants c(= 49.2 A. h = 47.6 A. (’ = RI.7 .A. /I = 109.5’. The crystals diffract t,o 2.5 A4 along the DNA axis and to at least, 3.0 A in the perpendicular directions. Data werr collected at 4°C’ using a Xentronics area detector and an Elliott CX-13 rotating anode X-ra> generator. The reflections were aut,oindexed and integrated with the on-line program suite BUDDHA (Mum uf al.. 1987). The data were reduced and merged using programs from the CCP4 suite (provided by 1’. R. Evans). Statistics for the data are shown in Table I.

(b) r)‘tructure

determination

The structure was solved by molecular replacement. The model of 434 Cro complexes with a 14 base-pair fragment (Wolberger et wl., 1988) was used as a starting search model. The fast rotation function (Crowther, 1972) showed 2 clear solutions, corresponding to the 2 possible orientat,ions of the complex around its pseudo X-fold axis. Both were used in translation R-factor searches (Bhuiya & Stanley, 1964). The solutions from both searches were virtually indistinguishable by R-factor. which was 43.8% to 3.0 A for the slightly better orientation. At this stage. the 20 base-pair model of DKA from the Rl-69 complex structure was exchanged for the 14 base-pair model. The R-factor remained almost unchanged (4459/h). To study the packing arrangement, the resulting model obtained was displayed using the program FRODO (Jones. 1978) on an Evans and Sutherland PS390 graphics station. The packing was reasonable, with the DKA molecules stacked to form a pseudo-continuous helix with overhanging ends paired. P;o unacceptable contact was found. The new model was refined with the program CORELS (Sussman et al., 1977), first treating it as a rigid body and then breaking it into smaller and smaller subunits. The resolution was increased from 5 to 2.8 x in several steps. The final R-factor to 38 A was 26!/,. The alternative orientation of the model was refined in the same way to a final R-factor of 28%. The better model was studied and rebuilt using FRODO. After rebuilding, it was refined with the TNT suite of programs of Tronroud et al. (1987). The model was rebuilt several more times and the resolu-

OR1

Figure 1. Sequence of the 20 base-pair fragment. The numbering follows the convention used by Aggarwal et al. (1988), from 1 through 7 in the 5’ to 3’ direction. Nucleotides in the opposite strand are primed. The central 14 base-pairs correspond to the true 0,l sequence.

Phage 434 Cro/O,l

323

Complex at 2.5 L! Resolution

Table 1 Data collection and refinement statistics Resolution

(A)

Number

Rs,, (%)t

267 532 694 831 911 895 899 884 780 539

7.1 8.1 90 103 11.4 12.5 16.8 21.1 280 31.7

7232

10.1

7.90 559 4.56 3.95 3.54 3.23 2.99 2.79 2%4 2.50 Overall f The R-factor

between symmetry-related

Shell (yb)Q

R P6) II

75 84 88 %H) 91 x9 87 84 80 1-L

75 89 93 94 94 83 77 70 59 38

18% 198 18.2 168 19.1 234 25% 290 31.0 30%

74

74

21.5

Sphere (%)I

reflections

is defined as:

Rrym = W-(WWil~ of an individual measurement and (I); is the mean value for all measurements

where Ii is the intensity of the reflection. 1 Percentage of reflections observed in that particular resolution sphere. 5 Percentage of reflections observed in that particular resolution shell. II The R-factor between the observed and calculated amplitudes from the refined model is defined as:

R = Z,IF-F$Z,IF;I. \vhew F, and F, are the observed and calculated

tion increased in several steps to 2.5 A. At this stage, the R-factor was 257o. Individual temperature factors were refined and 25 solvent molerules were assigned. The DIVA was regularized at several stages using the program CORE18 to ensure t’hat the sugars were in the C-2’ endo conformation. The side-chains of several residues were deleted from the calculations at more than I stage, and 2)FJ - IF,1 and IJ’J -IF,/ maps were studied to confirm their position. The same was done with the central bases in the operator. Tn all cases, the model showed good agreement with the density. Finally. maps phased with only the 14 base-pairs corresponding to t’he 0,l sequence were computed. The density for the remaining 6 base-pairs was clear and consistent wqth the model. The final R-factor for data between 10 and 2.5 A is 22%, using a model with 25 solvent molecules. The root-mean-square (r.m.s.t) error for the bond distances is 0.01 A, for torsion angles, 2.3”. and for the planar groups, 001 A. The average torsion angles for the DNA sugar-phosphate backbone are

s(= -35.7” fi=

140.1”

(54.4”). I= (46”).

1599”

(357”).

i: = - 157.2”

(13.8”)

(27.4”), and the average glycosidic

y = 53.0” (56.7”). and

t=

-128.7

t,orsion angle (x) is

- 111% (13.5”): standard deviations are shown in parentheses. The H-factor is shown as a function of resolution in

Table 1. The alternative orientation of the model was also refined to 2.5 A. The final model of the 1st solution was used as a starting point, but with the DKA in the opposite orientation. The model was extensively refined to an R-factor of 24%. At this stage, maps with 2lF,,I - IF,1 and

IJ’J - IB’J coefficients were computed. The maps, although calculated

with

phases

from

the 2nd model,

were

more

consistent with the 1st model. This indication was particularly clear at the ends of the DNA fragment. The DNA forms a pseudo-continuous helix; the positions of the “missing phosphate groups”, corresponding to the 5’ ends of each strand, are different from the 2 solutions.

A difference map caalculated with the phases from the 2nd t Ahbrrviation

used: r.m.s..

root-mean-square.

amplitudes.

respectively.

model showed has phosphate

clear positive peaks where t*he 1st model groups and the 2nd does not, and negative

peaks where the 2nd model has phosphat,e groups and the 1st does not’. This result, illustrated by Fig. 2. confirmed unambiguously the choice of the 1st solution.

3. Results (a) Overall structure of the complex The complex is shown in Figure 3. Each Cro monomer interacts principally with one half-site. The helix-turn-helix is positioned so that t’he amino t,erminus of helix 2 interacts with the DXA backbone on one side of the major groove and the turn and helix S/helix 4 loop interact with the backbone on t’he other side of the groove. Helices 4 and 5 cont’ribute to the dimer interface. Despite asymmetries in the central and flanking sequences, the DNA backbone is almost’ symmetric with respect to the pseudo dyad that relates one half-site to the other. The best least-squares superposition of sugar-phosphate backbone atoms in one half-site ont’o corresponding atoms in the other yields an r.m.s. difference in at,omic positions of 1.21 Lt and a rotation angle about the best axis of 180”. A surprising result is that when the protein monomers are superposed, the r.m.s. difference in atomic positions is 0.61 A, but the required rotation angle is 174.6”. Thus, the DNA backbone has a structural 2-fold, coincident with t’he pseudo 2-fold of the operator sequence, but t’he protein monomers are not related by this dyad. The asymmetry is not pronounced, but there are necessarily small differences in the way the two monomers inttaract with their half-sites. We have adopted the following numbering scheme to describe the structure. consistent with

Figure 2. Part of a difference Fourier map of the 2 possible orientations of the complex. The final model is shown as filled bonds: the alternative solution is shown as open bonds. The differencath map \?;a~ c,omputrd using tht, c+al(~ulatrtl amplitudes and phases of the 2nd solut,ion. Positive contours are shown as continuous lines and neant,ive caontours arc’ drawn with broken lines. The drawing shows the juncation between the 2 cryst,allographic~ally related molec&!s. The ia,& phosphate groups of the 2nd model are clearly m negative density. where the 1st model does not placse a phosphatr group. The phosphate groups of the correct solution arr in positive density. where the 2nd solution does not have a phosphate group. This map confirms unambiguously the c,hoice of the 1st solution.

t!hr numbering of t)he Rl-69/&l complex used 113; &garwal et al. (1988). Base-pairs on the left and right-hand sides of the operator are numbered 1 I, t,hrough 71, and 1K through 7R, respectively, from

the outside to the center of t)he sit?. Flanking basrpairs are - 11,. -81,, etc. Nucleotides on the complementary strand are given primed numbers (see Fig. 1). The 71 residues in each (ko tnonomer

Figure 3. Stereo drawing of the complex of 434 Cro with I)KA. The bent conformation of the DSA is apparent, as well as the extensive interactions all along the DKA--1)rotein interface. The top monomer corresponds to the R half-site. t,hr bottom to t,he T, half-site.

Phage 434 CrojOJ

3R

Complex

at 2.5 d Resolution

3R

4R

325

4R

-3R

3R b

I

Figure 4. Stereo center-of-mass plots for DNA backbone atoms. The cent’er of mass of each base-pair, excluding the base itself. is calculated

and plotted as an open ball. (a) Comparison of DKA in the 434 Cro complex (filled bonds) with standard, straight B-DNA (open bonds). It is clear from the plot that the DNA from the complex is more irregular and that it curves sharply at the ends. The central 14 base-pairs follow more closely a straight helical path. although local irregularities are clearly present. (b) Comparison of DNA conformation in the 434 Cro complex reported here (filled hods) w&h the Rl-69 complex (open bonds) described by Aggarwal et al. (1988).

to their are numbered from - 1 to 69, corresponding alignment’ with residues 1 to 69 of 434 repressor, and I, or R can be added to the residue number to denote the monomer bound to the 1, or R half-site. (h) DNA

conformation

The DNA forms a R-type helix, with a mean rise of 3.24 A per base step and a mean twist, of 36.1” per turn, corresponding to i.en bases per t,urn as in canonical B-DSA. All the sugars in our model have the C’-2’.endo conformation, because of constraints imposed during the refinement, but’ there is no evidence from difference maps at this resolution for any deviation from C-2’.endo. The r.m.s. values for the backbone angles are given above. The 5’-terminal bases in b0t.h strands are paired with bases in crystallographically related molecules, and the stacked fragments thus form a pseudocontinuous helix that runs along the unit-cell diagonal. The overview of the complex in Figure 3 shows that the D2iA is bent. The space curve in Figure 4, representing the locus of center of mass of the 19 base-pairs within a fragment, shows that curvature is concentrated at the ends of the operator. The central 14 base-pairs of the present struct#ure superpose well on the co-ordinates of the 14-mer in

complex wit’h 434 Cro (r.m.s. difference for backbone, 1.1 A). reported in a previous account, from this laboratory. Tn that paper (Wolberger et al., 1988), the DNA was described as relatively straight. Adjacent 14-mers in those crystals do. however, stack in such a way that’ they mimic rather accurately t,he curvature at the ends of the 20 base-pairs fragment (Fig. 5). Because the stacking varies somewhat a.t the six crystallographically distinct junctions, the extent. to which it reveals the curvature in a rontinuous fragment was not recognized in the earlier analysis (Wolberger et ab., 1988). Base-pairs in the operator are significantly noncoplanar (Table 2 and Fig. 6). In the Rl-69/0,1 complex, base-pair non-coplanarity was observed to give rise to a pat,tern of bifurcated hydrogenbonding (Aggarwal et al.. 1988). Non-Watson-Crick hydrogen bonds between adjacent bases on opposite strands might compensate for thr distortion of Watson-Crick hydrogen bonds when base-pairs are propeller twisted or buckled. We can assign at least one bifurcated hydrogen bond at t’hr center of OR1 in the Cro complex. The differences bet.ween the DNA backbone conformations near phosphate 4’ in the two complexes are such that, identical patterns of non-coplanarity are not likely. As shown in Figure 6, thymine 7’1, and adenine 6L are linked by a bifurcated hydrogen bond in the major groove.

1’R -1’R

Figure 5. Comparison of the DK‘A in the 2 complexes of 434 Cro with 11X.4. The 20 base-pair DSA from the X14 (‘ro complex is shown in thick lines. In thin lines is shown the 14 base-pair DSA and its 2 adjacent non-c:rystallogrsphic:ttllS symmetry-related neighbors, from the structure presented by Wolberger it al. (1988). The DKA in the 2 structures is very similar and bends in the same manner. The stacked DK’A molecules in t,he 14 base-pair structure follow rlosely the path of the 20 base-pair DNjA.

Less favorable minor-groove linkages can be considered for guanine 51, and thymine 4’L in the L half-site, and guanine 6’R and thymine 7R in t’he R8 half-site. Tn all three cases, large propeller twists of the base-pairs bring the releva.nt, groups into proximity.

(c)

Protein

conformation

The structure of free 434 Cro has been described (Mondragon et aZ., 19896). The protein consists of five cr-helices, which form a compact. bundle with a hydrophobic interior. Helices 2 and 3 form the Comparison of the motif. helix-turn-helix complexed and free proteins shows that there is no major change in conformation upon binding t’o

DXA. The principal changes occur in torsion angles of side-chains that lose or acquire important interactions on binding; i.e. side-chains that interact wit’h symmet’ry-related molecules in the crystals of’ free Cro and those that, contact DNA or help form the dimer interface. Tn helix 3. residues 27. 28, 29 and 32 change conformation and cont,act the DNA. Phenylalanine 46 changes conformation and fits into the dimer interface, and arginine 43 and glutamic acid 47 move t)o form a salt-bridge between monomers. In the structure of the free prot’ein, only the first 65 of t~he 71 residues are ordered. In the complex, one of the monomers (R) has 63 ordered residues and the other (T,) 66. The ordered residue at the carboxyl terminus of the L monomer is held in place by a crystallographic contact.

Phage 434 (‘rojO,l

327

at 2.5 A Resolution

Complex

Figure 6. Stereo view of the central base-pairs. The high degree of non-coplanarit? of the bases is apparent. giving rise t,o non-Watson-Crick hydrogen bonds (broken lines) between adjacent base-pairs. Similar networks have been observed in other structures (Nelson et al.. 1987; Aggarwal et al., 1988). The hydrogen-bond networks in the 434 C’ro complex and in the RI-69 complex are somewhat different. although the sequence and overall bending of the 0,l DNA is the same in the 2 comldexes.

(tf) Dimw

interfacr

The interface has a hydrophobic core, dominated by the bulky side-chains of phenylalanine 46 and tyrosine 61. and polar networks at its lateral margins. Residues in helices 4 and 5 and in the loop connecting helixes 3 and 4 contribute to the contact. The posit,ion of phenylalanine 46 is incompatible with a perfect, 2-fold. That is, the phenyl groups on

monomers related by the best Dr\‘A dyad would collide (Fig. 7). This interference probably cont’ributes to t,he generation of asymmetry. Arginine 41 and tyrosine 61 from one subunit and glut)amate 47 from the other form a small net,work of hydrogen bonds. Side-chains of these residues have different conformations on either side of the approximat’e dyad. although the patt,erns of polar bonding are similar (Fig. 7).

Table 2 Helical

-41~ A -31~ A.T -3’R -21~ (i.(‘ -YR -1RT.A -1’R 1K A,A I’R YR 21~ (‘.(i 31~ A. T 3’R 4K A.T 4’R 5K A.T 5’R 6K (‘,G K’R 7K ‘T’.A 7’R 7’11 I‘. A 71, 6’1,T.A 6L :i’l,(‘.(: 51, 4’11 ‘I’. A 411 3’1, 1’ A 31, 2’1. G’( PI, l’l,‘l‘~A IL - 1’1Ij 9.7’-1L -2’1,T.A -2L ‘1’ --3L

parameters

for the 20 base-pair

Twist (deg.)

Rise (A)

32.3 346 44.3 28.7 33.4 41.8 39.2 36.0 32.8 37.0 39.2 41.9 32.1 42.3 334 30.6 38.2 31.1

fragment

of DNA

Prop. t,wist wx.) 12.0 cl 100 8.1

136 12.X 26.7

13.0 17.1 143 200 29x %3

Buckle (deg.)

423 30 64 3.7 - 2.4 -132 - 84 -51

- 13.5 2.2 7.7 9.7

19.6

19.6 25.3

"3.2

6.1 43 44

4.9

14.1

X.1 -11.7 46 - 2.6

All the parameters were calculated from the final, refined model. The values of the rise and twist are affected by local variations in the structure, and thus serve only as a guide. Local twist and rise were c&ulatrd using the program HELIX; propeller twist and bucklr were calculated with the pro~r&*~ l%ROl,l, (Dickerson & Drew, 1981).

Figure 7. Stereo view of the protein dimer interface. The nebwork of polar interactions at the periphery of this interface is shown by broken lines. The arrangement of the monomers is not symmetrical. giving rise to different sidrchain interactions. The view is perpendicular to t,he DNA pseudo Z-fold axis: the DNA would run vertically along the left-hand side of the diagram. with arginine 43 projecting into the minor groove.

(e) L)iVA-Jl?YltP~Y~intrractiorrs (i) Sugar-phosphate

backbone

bind to segments of DNA in either side of the major groove. On t’he “unprimed” strand, a continuous ridge of atoms between C-3’ of nucleotide 2 and the phosphate group of nucleotide I is in van der Waals contact with the protein. We 1. Hydrogen bonds to phoscall this ridge segment phate groups - I and 1 appear to be of particular importance. The pept’ide nitrogen of glutamine 17. OY of threonine 16. and N” of arginine 10 all form hydrogen bonds with phosphate - 1. Tn the I, subunit, there appears to be a water-bridged hydrogen bond between thrronine 18 and phosphate - 1. The Cro monomers

side-chain of glutamine 17 donates a h@rogrn bond to phosphate 1. Glutamine 17, argminr 10 and glutamate 35 are all involved in a network of hydrogen bonds that help est’ablish the configurat,ion of the helix l/helix 2 corner appropriate for t’hr inter,actions with the DKA backbone (see Fig. 8). These interactions with DNA are also present in the 121-69 complex; the protein intramolecular interactions occur bot,h in the complexes and in the free proteins. On t,he “primed” strand, t,here is an even more extensive set of interactions, wit,h a region we call segment 2. The loop between helices 3 and 4 runs between phosphate groups 5’ and 6’ in such a wa) that the main-chain amide nit’rogrn atoms of lysint, 40 and argininr 41 donate hydrogen bonds to phos-

Figure 8. Stereo drawing of the interactions between Uh’A and the protein in the vicinity of base-pair I. The direction of view corresponds to Fig. 3, and this diagram is essentially an enlarged version of the bottom crntrr of’ that Figurts. Solvent molecules are labeled SOL. The amino end of hrlix 2 comes close to the phosphate backbone. and several hydrogen bonds link the phosphate groups and the protein. A water-mediated contact between threoninr 18 and 1 of the oxygen atoms at position - 1 is shown. The conta& between residues in helix 3 and adenine I are shown. The drawing shows the 1, half-site.

Phnge 434 ProjO,

Complex

at 2.5 d Kesolu~tom

329

T /

R

i

Figure 9. Stereo diagram of the protein-DX\;r\ interactions around the caent,rr of the operator. The loop connecting helices 3 and 4 comes close t’o the UK-A phosphate backbone and several hydrogen bonds join main-chain nitrogen atoms with phosphate oxvgrn atoms. Although the (‘ro dimer does not have strict %-fold s!-mmetry. its position in the operator and lwal small adjustments in the DNA backbone allow rssrntiall?. itlrntic*al interactions in hoth half-sitcts. phate 6’ and those of arginine 43 and pherqlalanine 44 donate hydrogen bonds t’o phosphate 5’. Both monomers make essentially identical contacts. These interactions are illustrated by Figure 9. The side-chain of arpinine 43 is inserted into the minor groove. but no direct hydrogen bond is made with the bases or thr backbone. The I>NA bacskbone lies so close to the protein backbone t’hat there are also extensive van tier Waals interactions all along the interfacca. The position of the phenylalanine 44 sidechain is particularly not’eworthy. It lies against’ t,he deoxyribose of nucleotidr 5’. This residue is also a phenylalaninr in 434 repressor. but the side-chain conformation is different, and therefore the nature of its c.ontac:t wit I-11)XA backbone. The (*ontact lies just’ in the interval where we find the largest shift bet’werrr t,hta I)SA conformations in the two complrxes. The difierenre in phenvlalanine conformation is present in the free protein (MondragStin it rrl.. 1989h). and we t,hcrefore suggest that it contributes to thr difierenc*e in l)KA backbone c,onformatiori brtwern c~ornplexrs. Another set. of c.ont,acts to segment 2 is made by

the turn bet)wren helices 2 and 3 and bv srrine 30. The main-chain a,mide nit,rogen atom o? residue 27 forms a hydrogen bond t,o phosphat’c> 4’. On the 1, half-sit’e, thr lysyl side-chain extends t’o form s saltbridgr with phosphate 3’: this interact,ion is not present ~II the K half-sit’e. Phosphate 4’ is hydrogenbonded t)o the 0’ of serine 30, the third residue of 3. These chontacts wit,h phosphate groups 3’ helix and 4’ are not present in the RI-69 complex, because residue 27 of RI-69 is thrconine and because the phosphate 4’ lies f&her away from the prot.rin. In the left half-site of the RI-69 complex. there is a water-mediated contact between phosphate 4’ and srrine 30. and the side-chain of serinr 30 form+ hjdrogtbn bonds with threonine $6 and threonine -7. These residues hwome valimx and Iysine. rtsspectivel?-, in 434 00. relrasitip Srririe 30 to turn aronnd towards phospha,te 4’. (ii)

Ntrar-puir.s

IZssrnt ially all the interact)ions with bases involve residues in helix 3, with the exception of van der \Vaalh contac+ from l$-sine 2i. Adenine 1 is hydro-

T

Figure 10. Stereo drawing of the specific interactions between residues in helix 3 and DNA. The L monomer helix 3 and base-pairs IL to 51, are shown. Note hydrogen bonds between adeninr 1 and glut’amine 28 and between guanine 2’ and glutamine 29. The bond to guanine 2’ is rather long in this half-site. but much more favorable in the right half-site (where a reasonable 2nd bond to N-7 can he assigned). Contacts wit)h the phosphate backbone aw made by the main-chain nitrogen atoms of residues 27 and 30 and by lysine 27. The side-clhain of leucine 33 lies cIose to thynine 4%. in clear \ran der LVaals contart.

apparent, when the t.wo monomers art’ supt~rpose(I. This is not t#hr case for the I)NA, whc~rc~ s~~gar phosphate hackbones of the two half-s&s are rather accurately related hy a &fold axis c*oinciderrt wit,h the operator pseudo 2-fold axis. This asyrnmt~trg~ in the protein arrangement, is morr apparent in the regions away from the DNA and in the sidcl-chains around the dimer interface. The contacts wit,h 1)X:4 are almost identical in hot)h half-sites, hecause of t hca way the asymmetric (lro dimrr lies against t.hti operator and because of some srna.11 Ioc~al adjust.ments in the DNA itself. Figure 11 shows t.hrL tlvo half-sites superposed on txach other hy t h(a Z-fold transformation relating thr 1)N.A. The protein monomers do not superposr as well as tlrc I)NA, although it, is quite clear that helix 3 in both monomers is in an almost identical position with respect to the 1)NA. The same is true of t’hr loop joining helices 3 and 4 and. to a lesser degrrr. of the amino terminus of helix 2. Some of the (nontacts between protein and 11X.4 are illustrated hy Figure t 2, wit.h the t.wo half-sites superposed. Tntera.ctions at the climer interface appear 10 generate the asymmetry. Bulky sirh-(*hains from phenylalanine 46 and tyrosine 61 o~~ul)y th(, intcrface. T’henylalanine 46 changes its conformation with respect to the frer protein ill or&r to tit. Moreover. it,s conformation is not identicsal in t tit1 t’wo monomers, since a symmrtric~ arrangrmf~nt of the two phenylalanine residues would lead to strric clashes in the interface. The n&work fortnvd hy arginine 41, glutamatr 4i and tyrosinfs 61 is also different in either side of the pseudo-dya,d. .Ipain. a perfectly symmet.ric arrangement of t hfw sidechains is strric~ally unfavorahlt

gen-bonded to glutamine 2X. The N-6 and S-7 nitogen atoms of the adenine form hydrogen bonds t)o the 0” and N” at,oms of the glutamine. The bonds are slightly long in our refined structure (R halfsite: K-6-W’ = 3.39 8; X-7-N”’ = 3.21 8. I, halfsite: N-6-W’ = 3.65 A: N-7-NE2 = 3.22 A) hut the fact that they are present also in the Rl-69 complex strongly supports the assignment. On the right, halfsit,e: O-6 of guanine 2’ accepts a hydrogen bond from 29. The correthe side-chain N”’ of glutamine sponding distance in the left half-site appears t,o he too long, however. Both glutamine 28 and glutamine 29 adopt unusual conformations that are probably partially imposed by the need to make contacts with base-pairs 1 and 2. In the case of glutamine 29, a hydrogen bond from the main-chain nitrogen at’om to the side-chain carbonyl group and van der Waals contacts with the methyl group of thymine 3’ further constrain its position. The\- have different conformations in free 434 Cro (Mondrag6n et al.. 1989a). Base-pairs 3 and 4 have no hydrogen honds with the protein. The methyl group of thymine 3’ inserts into a van der Waals pocket formed hy t.he side-chains of glutamine 29 and Iysine 27. The methyl group of thymine 4’ is in van der Waals c,ontact with the aide-chaitis of glutaminr PI) ((@). serine 30 (Oy): and leucine 33 (Cd). as well as with all atoms of the peptide backbone from (y” of glutaminr 29 to C”” of serine 30. The role of leucinr 33 is particularly interesting, as this residue is different in Rl-69 and Cro. We discuss its significance in t.he s&ion hrlow. AlI other residues forming contac:t,s wit,h USA bases are identical in the two protSeins. 10 illustrates t.he interactions ht~twecn Figure residues in helix 3 and DiGA.

4. Discussion (a) Dime~r conformation The perfect

monomers are not, related protein 2-fold axis. II\ small deviation (approx.

r

The DNA is bent with respect, to standa’rtl K-I)I\;A. The bent conformation of I)NA is necessary to allow the caont,acts between phosphatJe

by a 5’) is

I

R----

__-.--

1

Figure 11. Superposition of’the protein/DKu’A half-&es using the I)NA pseudo Z-fold transformation. The ft half-site is shown as filled bonds. the L half-site as open bonds. The protein monomers are not related by a perfect Z-fold transformation. and their fit is t,herefore less precise. As can be seen, helix 3 of both monomers lies in almost exactly the same position wit.h respect to DNA: most of the differences are in regions far away from the DNA and in th(l dimer interfacr.

Phage 434 Cro/O,l

Complex at 2.5 d Resolution

331

Figure 12. Stereo diagram of some of the interactions between protein and I)?u’A for both half-sites. The 2 half-sites were superposed using the DPU’Apseudo 2-fold transformation. Filled bonds correspond to the R half-site. open bonds to the I, half-site. As can be seen, helices 3 of both monomers align almost perfectly and interact almost identically with DNA. This is not the case of interactions near the amino terminus of helix 2, where small differences in the 2 monomers cause differencaes in the interactions. groups 1 and - I and the amino terminus

of helix 2. The bending is more apparent at the ends of the D1CA fragment. Figure 4 shows a plot of the center of mass of the base-pairs for the 20 base-pair fragment and for straight B-DNA with ten base-pairs per turn. It is clear that most of the bending occurs at the ends, although there are deviations throughout the fragment. The degree of bending is comparable t)o that observed in the Rl-69 complex with the same DNA fragment (Aggarwal et al., 1988) and to the bending that can be recognized in the Rl-69 and Clro complexes with a shorter piece of DNA (Anderson et al.. 1987: Wolberger et al., 1988). In the latt)er structure. the bending was not originally noticed because of the shortness of the DNA and the variation in structure arising from the unusual packing arrangements. E’evertheless, when the two Cro complexes are compared, it is clear that they have similar bent DNA conformations (see Fig. 5). Gel retardation measurements have been used to assess DNA bending by 434 repressor and Cro in Recent experiments (G. Koudelka, solution. personal communication) show equivalent bending by the two proteins and require revision of the earlier interpretation (Koudelka et ~2.. 1988) that 434 Cro bends DSA much less than does repressor. The new experimental measurement’s are consistent with t,he cryst’allographic results reported here. (ii)

Width of mc~jor and minor

grooves

The minor groove varies in width. It’ is narrower at the center of the operat,or and wider at the ends. The width has a low value of 9.3 A betweeen phosphat,e groups 71, and 3% and a mean of 10.4 a for the central four base-pairs; its highest value, near the ends of the operator, is 13.7 A. The overall mean of 11.4 A is in good agreement with the expected value for B-DNA (11.5 .&: Saenger, 1984). The

narrowing of t’he minor groove at the center of the operator is probably stabilized by the two arginine 43 side-chains, which project between the phosphate backbones. The major groove is more regular. The variations in its width are all within 1 A of t,he mean value of 17.62 B; the expected value for standard B-DNA is 17.5 A (S aenger, 1984). The minor groove variations for the Cro complex are less marked than for the Rl-69 complex, which shows an even larger et al.: narrowing of the minor groove (Aggarwal 1988). This difference is due to a relative shift in the position of phosphate 4’. In the Cro complex, this phosphate lies near the helix S/helix 3 t,urn and forms hydrogen bonds with the peptide nitrogen atom of lysine 27 and the Oy of serine 30. In the Rl-69 complex. it shifts about 1 to 2 ,A into the minor groove and away from the protein. (iii)

Raw-pair

non-coplanarity

The base-pairs at the center of the operator show large deviations from planarity (Fig. 6). One consequence is the presence of bifurcated three-centered hydrogen bonds in adjacent pairs. Such threecentered bonds ha,ve been observed in other structures having t’hree or more adenine residues in sequence (Nelson et al., 1987; Aggarwal et al., 1988). More recently, they have also been observed in GC tracts (Timsit et al., 1989), showing that it is possible to have significant non-coplanarity in nonAT-rich sequences. In all cases, a large propeller twist is present in the base-pairs involved. We have considered three possible bifurcated hydrogen bonds in the Cro/O,l complex. One, between thymine 7’1, and adenine 6L, joins groups in the major groove; the other two, of less favorable geometry, involve guanine 6’R/thymine 7R and guanine 5L/thymine 4’L and join groups in the minor groove. The central base-pairs in the Rl-69 complex also form bifur-

(c)

1)9.4

-protrirl

interactiona

(lonserved atld variable base-pairs are segregated in the 434 operat,ors. The outer four base-pairs of the 12 nat,urally occurring half-sit,es are near13 invariant. Their sequence is S’-A(!AA. rxcept in thr right’ half-site of OR3, where G replaces A at position 4. The central six base-pairs of each sitmeare significaantly more variable: no strong consensus can ~JP discerned. although the sequencaes are relatively ATrich. The structures of Rl-69 and 434 (Ire in c*omplex with 0,l show that nearly all base-pair c*ontact’s involve conserved operat,or posit’ions 1 to 4. All contacts arc very similar. except for a single major groove van tier Waals cont.act t.0 base-pair 5 and minor groove water-mediated hydrogen bonds IO base-pair 7 in the RI-69 complex. In neither complex ark the cbentral four base-pairs in van clt~ Waals contact with any atoms in the proteitt. In discussing operator recognition by Kl-69, Aggarwal rt (~1.(1988) described the way cont)acts to base-pairs 1 t,o 4 can specify AC’AA uniqueIT, and t.heJ proposed that t,he central base-palm influence repressor afinity by affecting the relative ea.sr with which the I)NA backbone in this region ca.n adopt thr (sonformation itnposed by t,hr protein. In out prwent a,nalysis of recognition by 434 i’ro. WP sirnilarly distinguish between the role of direct contacts at conserved (or nearly conserved) portions and indirect effects at variable central positions. Wf note t,hat the importanc:cJ of DNA conformation in supplying a surface c:omplementar?to the protein necessarily couples these “direct” and “indirect” aspects, but the segregation of conserved and variablr base-pairs makes t.he dist.inc%ion useful.

Adenine 1 forms hydrogen bonds to glutamine 28: guanine 2’ forms a hydrogen bond to glutamine 29. Some of t,he hydrogen-bonding distances are rather long. but we note that t’hey are present in complexes of Rl-69 with C&l and C&3. as well as in the (‘ro/O,l structure described here. Moreover. an?; base-pair substitutions would lead to substantially less favorahle cortt,acts, with full?; unsat,isfied hydrogen-bond donors. In Rl-69 complexes. t,wo hydrogen bonds are found between glut.amine 29 and guanine 2, but only one is found here. The sidecahain conformations of glutamine 28 and glutamine 29 are critical for correct hydrogen bonding and van der Waals contacts; both are different from what is observed in free 434 Cro. They are held in place by hydrogen bonds to other groups in the protein: glutamine 28 to the side-chain carborryl group of residue 17 and glutamine 29 t’o its own peptidr nitrogen at,om. The multiple hydrogen-bonding caapacity of glut,amine is therefore essential for generating the networks found in this and the Rl-69

c~~nlpl<~xc~h.\\‘ithout thesca snpportittg int(1r;lc.t ion>, tJ> glt~li~ttiitic~ the partic~ulict~ c~ot~fortrralicJtis ;idol~i~cI 28 iI,tl(l plutatrtinc “9 \Voultl lJt’ot~:~tJl~~tW Iltlf;~vot~ ;tblr. Rc*c~ogttitiott ih thtls ii c~onc~t~rtc~tllJt’ol)+‘t’t>. ot’ these sitlc-(*hains and of their c~ttvirottttrc~trl u-it I-tin 1h? sul)utlit (SW also l’atjo Al (11.. IWO). Thc~rc ilt’(’ tlo ttytlrog~~tr t)otttls to l)ilhCs-lJitit’ 1). hIIt the methyl group of thyntittcb :!’ fits IlCittI,L’ into il. Viltl der LVaals pockt>t formr>d It!. the sith-chaitt of’ I?sittcL 27 a,ntl glut,atnint? 29. This interacat iott ~~J~wI’F. to specify A-T i1t positiott 3. Ylowo~er. iI st rrmgly cbonstrains glutatnine 19. c~orrelating rt~cogttitiott 01’ has+pairs 2 and 3. Thta IJropelIrr twists of bastb-pairs 3 anti 1 art’ quite laryth and somewhat different in t ttra Ipft and right half-sitcls. These tJ;t.st,-pairs ilr(’ also signi cantI\, kJuc*kletl. The edges of the bascas that c,cttttact the atnittcj ac*id sidt,-chaitls arc’ nottrthrrlcss in vet’>. similar position in both half-sit,tls with rrspc~~t to thcx prot>eitl, as (*at1 IJ~Lseen in Pigurf* Il. The tliffrrrttc:~~s in propellc~r twist b~~tw-rt~tt th(a two half-sittis I het,(lfore appt’ars to c*ompensattJ for small effects of’ prot,eiti dinicbr asytnmr,try and fijr small difit~rc~ti~*c~sin backbonr c.onformatiott. (?)mparisotr of t h(J 0, t complt~xes of 434 (‘ro and RI-69 shows l)h;rt tnajor groove caonta& involving side-chains of rcsitlnt~s 27 through 29 specif\- the c.hoicacbof c*onst~rvetl bast’pairs 1 through 3 in hot h c:ases. That is, tlespitr, thr, different I)SA c~onformations in thrs (+omptc~uc~satttl despite the asymmetry of’ t.tw t !ro tlitrttkr. itttrt,actions tvit.h t.he c.ompletLely invariattt lJar( c~f’ t.ht> opera1 or arv rsst%ntially c~cjns~~rvrcl.

Position

for the differential function of sinw it is the site of the ktt> non-consensus substit~ut,ion in 0,:~. The atiniLy of (“r.0 is relatively unimpaired hy this substitution ((: .(’ fat, A .T). whereas repressor binding is substantially decreased. Thrtar rrsidurs of (‘ro are in van drr \Yaals contact with the thyminr of basepair 4 in (),I : gtut,aminr 29. serinr 30 and ktrcitrf~ 33. All three contact the thymine methyl group. A cytosine base can be accommodated readily at position 4’: hut at the cost, of a “hole” where the thymitte methyl group would project. lVhy is this hole not unfavorable! One possibility. suggt~strd I)>. Wolberger rt crl. (1988). is t,hat the methyl group contacts arr* actually strainecl. sincse the dist.ances to peptide-group at,orns at the 29/30 peptide bond arc‘ relatively short,. Elimination of t.ht: tight caontacts when (: (’ is substituted for A .‘I‘ might comppnsatr for loss of’ van der LVaals interac6ons from t ht. thymine methyl group. X high-rc~sotutiorr strucatnrr, of 0,3 czomplexed with C’ro. now in &Jmgrwi. should answer this question. Whatever the full (~xplanat,iorr. WC‘ can rule out thr simple virw t)hat Irucirtr (residue 33 in 434 (Ire) “recognizes” T and (’ tqually at posit ion 4’ and that glutamine (residue 34 in 434 repressor) specifies T uniquely. Suhstit)ut,iott of Ieucinr for glutamine in 434 repressor does not dtrt its strotig prrferencbfk for A “I’ at, base-pair 4; substitution of glutamine for teucine in 434 (‘1.0 xc*tuall>

434

Cro

and

4 is critical repressor,

Phage 434 CrojO,l

Complex at 2.5 A Resolution

for G.C (G. Koudelka. creates a “preference” R. Wharton & $1. Ptashne. personal communication). In the present structure, C6 of leucine 33 lies in the crevice between the methyl group and O-6 of the thymine 4’. A glutamine extended as in the RI-69/(&l complex (Aggarwal et al., 1988) could not be inserted without some re-arrangement. Thus, even in two complexes as closely related as those of 434 repressor and Cro with O,l, the base-pair specificit,y implied by the identity of residue 33 depends criticall: on t,he precise local geometry of t.he prot,ein/I>hA interface. Leucine at position 33 in 434 Cro is not equivalent to leucine at position 33 in 434 repressor. Moreover, the contacts that set up the precise local geometry come from a significant part of the (!ro and Rl-69 molecules. These include the J)NA-backbone interactions described above, which involve residues between 10 and 44, as well as the major groove int’eractions involving helix 3. Thus: differences between the two proteins distributed over a substantial part of their structures appear to have c~onsequenc~esfor differential recognition at base-pair 4. (iii) C’wtral base-pairs There is no major groove contact to base-pairs 5, 6 or 7. The only minor groove contacts are watermediated and involve bonds from the guanidinium group of arpinine 43R to 7’L: 7’R and 6’R (Fig. 9). Both arginine 43 residues are in clear van der Waats contact with fhe sugars of these base-pairs and probably provide a stabilizing electrostatic field. Binding of 434 (‘ro responds to central base-pairs changes in essentially the same way as does Rl-69 (Koudrlka e:tul.. 1987): G. C and C. G base-pairs are unfavorable at position 7, and somewhat unfavorable at position 6. The arginine 43 interaction cannot readily accaount for the observations in the case of 434 repressor, since conversion of this residue t,o alanine does not change the pattern of repressor preference (Koudelka et a,Z.:1987). We propose that, as wit’h 434 repressor. t’he central base-pairs influence t,he affinity of 434 Cro by affecting the capacity of an operator to adopt the conformation imposed by the protein, rather than by altering a direct non-covalent contact. What a.sp
effects of sequence on a specific, local conformation. There are a number of ways in which a sequence of, say, ten base-pairs can generate approximately the same net twist or bend. By contrast, the conformation at the renter of ORI is constrained by the requirement that, it has a particular t,otal twist and, more importantly, by a series of interactions that precisely determine the positions of phosphate groups 4’, 5’ and 6’. and that’ also bear on deoxyriboses 5’ and 6’. Tt is the influence of the sequence on t.he free energy of the actual conformation. rather t,han the sequence dependence of the larger ensemble of all conformations with t,he same net twist or bend, that we must understand. In analyzing the central base-pairs in the Rl-69/O,] complex. Aggarwat et al. (1988) called attention to base-pair non-coplanarity in the bifurcated hydrogen bonds that can result. Do these hydrogfbn bonds. of admittedly rather marginal geometry, make measurable contribut’ions t,o the overall free energy of the operator, or is optimization of base-pair stacking a quantitat.ively more significant effect! The present. structure does not permit a de6nit.e distinction. It does show base-pair geornetries that “take advantage” of possibilities for forming bifurcated hydrogell bonds. These are different from the three-center bonds found in t’he RI-69/O,] complex. probably reflecting a difference in the operat,or backbone conformation near phosphate 4’. The observation that Rl-69 and 434 (:ro exhibit, a similar dependence of binding on central base-pair sequence suggests that operator adjust,merits to improve stacking may be at’ least as important as specific hydrogen-bonding. In conclusion. two general points emerge from analysis of t,his structure and from comparison with the 434 repressor complex. (1) Int,erac%ions with invariant, base-pairs are conserved. The two proteins dock similarly on the operator and recognize base-pairs l t’hrough 3 in the same way, by contacts from conserved residues. The small differences that, we do observe. for example in distances and inferred strengths for hydrogen bonds, do not contradict t’hr basic notion that there are essentially identical mechanisms for discriminating against substitutions at base pairs I through 3. (2) Differential recognition at the critical position 4 in the operator involves more t.han a local substitution of amino acid residues in helix 3. Contacts at this base-pair depend on differences in I>NA conformation brtwtlen the two eomplexrs. and the), therefore involve widely distributed amino acid substit’utions. The phage 434 genetic switch thus appears to involl-e straightforwardly conserved interactions for ident’ical sp&icitirs. but’ unanticipatedly distributed differences for divergent speGfic+ics. Tn t,hr case of repressor and Cro from bac*tcriophage 1. there has been a long-standing debate. caoncerning thr validity of point (1): do conserved residues make conserved cont,actsY (Hochschild &. Ptashne. 1986; Ohlrndnrf pt al.. 1982). The L regulatory proteins are similar only in their helix-turn-helix regions, and the clivrrgrnc
binding domains could. in principle. lead to different modes of docking on DNA. A direct (nomparison of repressor and Cro complexes from /1 is not yet possible, due to the absence of a high-resolution /1 Cro/operator struct’ure. We suggest, however, that our observations on the homologous 434 prot,eins add a clear precedent to the evidence favoring vonservrd interactions for the common rrcognit,ion pat.terns. We thank Marie Drottar for her essential role in crystallization and purificat,ion of protein and D?u’A. \Vr t,hank M. Ptashnr for continued int,rrest and collaboM. Hlum. (:. Koudelka. rat,ion. and A. hggarwal. I). Rodgers, and C!. Wolberger for help and suggestions. This work was supported by NIH grant, GM-29109 (to M. Ptashne). A.M. was a post-doctoral fellow of the I)amon Runyon- Walt’er WinchelI Cancer Researclh Fund (DRC: 878).

References Aggarwal. A. K.. R,odgers, D. W., ‘Drottar. M.. Ptashne, M. & Harrison. S. (‘. (1988). Recognition of a DSA operator by the repressor of phage 434: a view at, high resolution. Science, 242, 899-907. Anderson. tJ. E.. Bashne, M. $ Harrison. S. (‘. (1987). St,ructure of the repressor-operator complex ot bact,eriophagc 434. Nature (London), 326, 846~85P. Bhuiya, A. K. & Stanley. E. (1964). Molecular location from minimum residual calculation. Acta f’ystalloyr. 17. 746-748.

Hlum. M.. Metcalf, P., Harrison. S. (‘. XL Wiley. 1). ('. (1987). A system for collection and on-lint of X-ray diffraction data from a integration (‘rystallogr. 20. multiwire area detector. .I. Appl. ‘L35-242. C.!rowther. R. A. (197%). The fast. rotation function. In Thr MokcuEar Repkzcement ;M&od (Rossman. MI. (i.. ed.). pp. 173-183. Gordon and Breach, Xew York. IXckerson, R. & Drew. H. R. (1981). Strurt,urv of a R-DNA dodecamer II. Influence of base sequence on helix structure. J. Xol. Riol. 149, 76lb786. Fujimoto, R. S. & Schurr, J. M. (1990). Dependence of the torsional rigidity of DNA on base composition. ,Vature (London), 344. 175-178. Hochschild, A. 8: Ptashne, MM.(1986). Homologous interactions of i repressor and A Cro with the 1 operator. (‘cll, 44. 925-933. Jones, T. A. (1978). A graphics model building and refinement syst.em for mac~romolrcules. d. Appl. ~‘Tystaltogr. 11, 268-274. *Jordan, S. R., Whitcombe, T. V.. Berg, .l. 11. &. Pabo. t’. 0. (198Fi). Systematic variation in DNA lrngt,h

Koudelka. (;. K.. Harrison. S. (‘. & l’tashnc. &I. (l!#i). Effect of non-contacted basrs on the affinit! of’ 434 operator for 434 repressor and C’ro. S~trrr+ (I,ondorr I, 326, X86-888. Koudelka. (:. I<.. Harbury. I’.. Harrison. S. C’. B I’tashnr. $1. (1988). l)iYA twist,ing and the affinit,y of bactcariophage 434 operator for bacteriophage 434 rrprrssor. Proc. .\:at. Acd. Sri., I ’ S .A. 85. 4633 ~4637. Mondra,g6rr. .\.. Subbiah. S.. itlmo. S. 1’.. I)rott,ar.. M. & Harrison. S. C’. (1989). Struc%urr of t III. aminoterminal domain of lhagtz 43-C rrprtvisor at 2.0 S resolution. ,/. Xol. Kiol. 205, 18!)&‘LOO. Mondrag611. A.. \Volbergc>r. (‘. L+ Harrisotl. S. ( ‘. (I!)%~). Struc+urv of phagv X1-C (‘ro 1)rott:irl at ?,3S:f rtlsolut,ion. ./. .Mol. Niol. 205. l7!)-I X8. Nelson. H. C’.. Finc.h. .J. T.. Luisi. 13.F. Nr Klug, .\. (1987). The structure of an oligo(dl\) oligo(dT) tract, and it)s itn1)lications. .v/ll/rre /Lnndon). 330. biological “2 ] -226. \c’. P., Fisher. I<. (j.. ‘l‘akrda. Ohlendorf. I). H., Antiwsoll. Y. Cy:Matthew-s. IS. \V. (19x1). The mo1evuta.r basis c)t DNA protein recognition inferred from t,he struvturv , 298. 7 I H-73%. of ( ‘1.0relbressor. Nnfvr~ / LorrAon) Pabo. (‘. c).. ,!ggarwal. .\. K.. vJortlan. S. I<.. J%ramvr. I>. .I.~ Obrysc~kar~. I’. R. Pr Harrison. S (’ ll!190). (‘onservvd rrsidurs make similar c,otttac*ts iti two rt,I’ressor-oprrator c~omplexrs. Sci~ncr. 247. lZl0 l%l3. Ptashnr. hl. ( 19%). il C&n~tic Sulitch. (‘ell f’rpss and Rlac.kwell Scientific Publicaations. (‘ambridge, ,\I~‘\. Saenper. L\:. ( 1984). rrinciples 0f .Tuclaic Acid ,Ttr~ct~lre. Springer-Verlag. New York. Sussman. .L .l. I,.. Holbrook. S. R., (‘hurrah. (:. 31. & Kim. S ( 1977). r\ structure fa<.tor least,-squares rt~linrmt*nt proctdurr for macromolec~ular structurf~s tlsing VOIIstrained and restrained parametcsrs. .4ctcl ( 'rptalhyr. wrl. .A. 33, 800-804. Timsit,. \‘.. LVesthof. E.. Fuc.hs. Ii. I’. I’. & b1ora.s. 1). (l!jX!t). L‘nusual helical packing in cbrystals of I)S;\ bearing a mutation hot spot.. ,Vntu~, / /,onr/on/ 341, 4X-46P. Tronroud. I). E.. Ten F:yc.k. I,. P. A Matthrws. I<. \V. (1987). An rfficient general-purposr Irast-squarva refinement program for macromolec~ular struc$urr. Acts (‘r+stallogr.

.src’t. .-I. 43. 489-501.

Wharton. R. I’.. Hrowvn. E. I,. & Ptashne. M. (1984). Subst)ituting an a-helix switvhrs t)hr s~c~u~:nc,r-sprc,it(~ 1)S.i irrtrravtions of’ a repressor. (‘rll. 38. 361 -: