Proc. Nat. Acad. Sci. USA Vol. 71, No. 3, pp. 593-597, March 1974

The lac : Molecular Shape, Subunit Structure, and Proposed Model for Operator Interaction Based on Structural Studies of Microcrystals (electron microscopy/x-ray diffraction/protein-DNA interaction)

THOMAS A. STEITZ, TIMOTHY J. RICHMOND, DAVID WISE, AND DONALD ENGELMAN Department of Molecular Biophysics and Biochemistry, Yale University, New Haven, Connecticut 06520 Communicated by Frederic M. Richards, October 31, 1973

ABSTRACT Electron microscopic and powder x-ray We would like to suggest another model for repressor- diffraction studies of small crystals of the pro- tein provide evidence on its molecular shape and subunit operator interaction that is consistent with the asymmetric structure which in turn suggests a possible mode of repres- molecular shape of the repressor derived from x-ray and elec- sor-operator interaction. The crystals are probably ortho- tron microscopic studies reported here. This model allows all rhombic space group P2221 with unit cell dimensions of four subunits of the repressor to interact with the operator. a = 140, b = 91, c = 117 A. This tetrameric protein appears It proposes 222 symmetry for the repressor, as has been found rather asymmetric, having approximate molecular dimen- for all sions of 140 A by 60 A by 45 A. The dumbbell shape of the other tetrameric of known structure (16). projected molecular outline observed in the electron Finally, it is consistent with the partial 2-fold axis found re- micrographs can be explained by assuming that the sub- lating the nucleotide sequence of the operator by Gilbert and units are related by 222 symmetry and are placed at the coworkers (13). corners of a plane rectangle. We propose a model for re- pressor- operator interaction in which the DNA binds to Crystallization the repressor with its long axis aligned with that of the repressor and with its 2-fold axis coincident with a twofold The lac repressor protein used in these experiments was iso- axis of the repressor. lated at Harvard in Prof. K. Weber's laboratory by Dr. T. Platt and TAS (14). Crystals were obtained by dialyzing The control of DNA into messenger RNA is best repressor protein against 40 mMNl phosphate, pH 7.0, and understood in the case of the lactose of 10 1iNI isopropyl-f-D-thiogalactoside. The crystals are thin (1-7). Transcription of the lactose operon by RNA needles up to 0.2-mm long. Although the crystals obtained is prevented by the specific binding of the lac repressor protein thus far are too small for single crystal x-ray diffraction (150,000 molecular weight) to the operator locus. Induction analysis, we have obtained electron micrographs of negatively of 0-galactosidase by , such as isopropyl-f-D-thio- stained crystals (Figs. 1 and 2) and powder x-ray diffraction galactoside, has been explained by the binding of the photographs (Fig. 4). to the repressor which causes the repressor to dissociate from the DNA allowing transcription to proceed (1, 3, 6). We have Electron microscopic studies initiated structural studies on the lac repressor protein to From negatively stained electron micrographs of these re- elucidate the nature of the ol)erator-repressor interaction pressor crystals (Figs. 1 and 2) we have determined two unit- as well as the allosteric transition produced by inducer cell dimensions, the lattice symmetry of one projection and binding. the approximate size of two dimensions of the tetramer. In The interaction of the tetrameric lac repressor protein with the most ordered part of the electron micrograph, as judged DNA poses some interesting structural problems. Does the by the resolution of the optical diffraction pattern, two unit- repressor recognize double-stranded DNA (8, 9)? Does it cell dimensions were measured as 95 A and 115 A. Possible locally denature the operator and bind to single-stranded DNA errors in magnification and, thus, in cell dimensions are esti- (10)? Or does it bind to a region of the DNA that assumes a mated to be + 10%. The crystallographic symmetry elements cloverleaf structure (11, 12)? At least two models of repressor- observed are a glide line parallel to the needle axis, 2-fold operator interaction have been proposed to explain how the axes perpendicular to the micrograph, and mirror lines per- four identical subunits of the repressor, which has four binding pendicular to the long axis of the crystal. Hence, the symmetry sites for inducer (3), can bind to one operator and make use of this projection is probably pgm, which suggests that the of all four subunits. Muiller-Hill and colleagues (9) suggest minimum crystal point group symmetry is D2. The simplest that the four subunits are arranged in a superhelix so that interpretation of the micrograph is that each tetrameric each subunit can bind to a DNA sequence that is repeated molecule appears in this projection as a dumbbell or waisted four times in the operator. Sobell (12), expanding on a model ellipsoid (Fig. 1 right; Fig. 2). Although these dumbbell shapes proposed by Gierer (11), proposes that the operator folds intq might result from a more complex superposition of molecules, a cloverleaf structure that possesses approximate 4-fold sym- we have not found another interpretation consistent with all metry relating its arms, and that the repressor likewise has the data. a 4-fold axis relating its four subunits, each of which contains Optical diffraction patterns (Fig. 3) of electron micrographs an identical DNA binding site. of these crystals show reflections out to 15 A along the crystal 593 Downloaded by guest on September 28, 2021 594 Biophysics: Steitz et al. Proc. Nat. Acad. Sci. USA 71 (1974)

FIG. 1. Electron micrographs of lac repressor crystals. (left) An electron micrograph of a negatively stained lac repressor crystal taken using an Hitachi HU-11E at about 51,000 times magnification at 100 kV. The crystal was crosslinked with 0.1% glutaraldehyde, placed on a carbon-coated Formvar film grid, and stained with 0.2% uranyl acetate. Square encloses one unit cell containing four molecules as shown in the right of Fig. 5. (right) A partial real-space photographic average of the micrograph shown at the left. This real- space averaging of the projected contents of several unit cells was accomplished by very precise of the print by exact multiples of the unit-cell vector parallel to the long crystal axis and taking a multiple exposure photograph through a photographic enlarger, a method used by Labaw and Davies (20). Four to six translations produced an image that could not be improved by additional averaging. The most interesting part of the micrograph is to the left, where the unit-cell boundaries are clearest, suggesting we are viewing down the long a axis of the crystal. In this part of the photograph one can see white dumbbell- or waisted ellipsoid-shaped objects each of which corresponds to the projected density of one tetrameric lac repressor molecule and has dimensions of about 45 i by 60 A. The b axis is horizontal and the c axis is vertical in these micrographs. Note that the pgm symmetry relating the dumbbell-shaped molecules is clearly visible, suggesting that the structural detail was not introduced by the translational averaging. The part of the crystal to the right which shows less detail is thought to be twisted, giving a view oblique to the a axis.

axis and 20 A0 across the crystal. The glide line results in in part on the two-cell dimensions obtained from the electron extinction of reflections 1 = 2n + 1. While the diffraction micrograph, the weak 91-A powder reflection is indexed as patterns of different areas of the micrograph vary in resolu- the (010) reflection and the strong 59-A spacing is indexed as tion and intensity distribution, those patterns extending to the (002) reflection; the (002) is also a strong reflection in the the highest resolution have quite similar intensity distribu- optical diffraction pattern (Fig. 3). tions. Unit cell dimensions, crystal packing, X-ray powder diffraction and molecular shape The most important result from the powder diffraction pattern Combining these data from electron micrographs and x-ray is that the largest spacing in these crystals is 140 A (Fig. 4). diffraction photographs, we conclude that the unit cell dimen- This reflection at 140 A spacing is best indexed as the (100) sions are 140 A by 91 A by 117 A with a = = y = 90°. reflection; attempts to index this reflection as the (200), These cell parameters yield a unit-cell volume of 1.5 X 106 A3. (110), (101), etc. predict reflections in the 140- to 90-A spacing If the unit cell contains four tetrameric molecules as the micro- regions which are not in fact observed. Thus, the third unit- graphs suggest, then the cell would also contain about 50% cell dimension, which is not observed in the electron micro- solvent or have a Vm of 2.5 A3 per dalton, the average value graph, is 140 A. Other peaks are observed in the powder pattern observed for protein crystals (15). at 90.8 A, 69.5 A, 58.8 A, 48.6 A, 39.2 A, and at higher angles. Since all three cell dimensions are different and since the Indexing of these peaks is not completely unambiguous be- micrograph shows a minimum symmetry of three perpendicu- cause of overlap of expected reflections. However, relying lar 2-fold (or 2-fold screw) axes, an orthorhombic crystal class Downloaded by guest on September 28, 2021 Proc. Nat. -Acad. Sci. USA 71 (1974) Structural Studies of lac Repressor Crystals 595

140 A

r- c6 69.5 A- f"qI\ 9-4 4 58.8 A 1.2 A 21

0 "0.005 0.01 0.015 0.02 0.025 0.03 1/d (k') FIG. 4. Radial 'intensity distribution of powder x-ray diffrac- tion patterns from repressor crystals. Powder x-ray diffraction patterns were obtained from crystals tightly packed into a 1.0-mm FIG. 2. An enlargement of a portion of the translationally sealed glass capillary and photographed at 40C in an evacuated averaged electron micrograph with our interpretation of the point focusing camera of the Franks type. A 65-hr exposure at a molecular outline drawn on the right. Each dumbbell is one 7.8-cm specimen-to-film distance showed at least seven measur- repressor tetramer and there are four tetramers per unit cell. able diffraction rings. At a specimen-to-film distance of 12.5 cm, the longest spacing observed was 140 A, although spacings to is indicated. The cell is probably primitive, and the 140-A 350 A could have been observed. To obtain the radial intensity spacing must be the (100) rather than the (200) reflection. distribution shown, a portion of the film was measured with Thus, the most likely space group is P2221. an Optronics drum scanner densitometer at 25-/Am intervals. Densities at equal radii from the center were averaged at radial intervals of 25 ,um, and the resulting intensities were corrected for background by a seventh order regression on seven selected background values.

Molecules can be easily packed into this cell (Fig. 5) in a manner that accounts for the stain distribution observed in the electron micrograph (Figs. 1 right and 2). Packed in this way, each of the four molecules extends the full length of the cell along a, and thus are stacked directly on top of one an- other when viewed along a, as in the electron micrographs. Such vertical stacks of protein in crystals that may be more than one unit-cell thick would exclude stain and account for

-4 y b 4

Fl

4- H 4

C 4. H 140 A ^ W 91& - FIG. 5. Molecular packing in two projections. (left) The [0101 projection, which was not observed in the micrographs. FIG. 3. Optical diffraction pattern of a portion of the electron We conclude nevertheless that the molecules extend the full micrograph shown in Fig. 1. The micrograph was masked to include length of the 140-A cell. (right) The [100] projection, which cor- the regions showing the most detail. The highest order reflections responds to the projection view in the electron jnicrograph (Figs. visible (though not on this photograph) are at 15-A resolution. 1 and 2). The dumbbell-shaped molecular outline observed in The b* axis is horizontal and the c* axis is vertical. An extinction the micrograph is indicated. Note the similarity to the model in for reflections I = 2n + 1 is apparent. Fig. 6c'. Downloaded by guest on September 28, 2021 596 Biophysics: Steitz et al. Proc. Nat. Acad. Sci. USA 71 (1974)

.-55-60oA - p- 45A-.i or C4 (a molecular 4-fold axis). A molecular 4-fold axis is ruled out since all three molecular dimensions are different. Thus, the symmetry relating the repressor subunits is likely to be D2 (that is, 222), again as found with all other known examples of tetrameric proteins. If the repressor's symmetry is D2, the possible arrangements for the subunits are at the corners of a tetrahedron, at the corners of a plane rectangle, or some. in- termediate to these arrangements. The dumbbell-shaped profile (Figs. 1 right, 2, 5 right) can most easily be accounted for by an arrangement (Fig. 6c) of subunits that is more nearly rectangular planar than tetrahedral (Fig. 6c). Possible models of repressor-operator DNA interaction The asymmetric and waisted shape of the repressor protein a. FRONT b. SIDE also places some limitations on models of repressor-operator interaction, if we assume that no major rearrangement of 55-60A quaternary structure occurs upon binding to. DNA. The model suggested by Sobell (12) is not likely to be correct, since it requires that the repressor have a molecular 4-fold axis, ~T which is not consistent with our data. The model advanced -4 5-50A by Muller-Hill and his colleagues (9) requires the subunits 1 of repressor protein to be arranged in a helical fashion around the DNA, which is proposed to consist of four repeating se- C. TOP quences. If the repressor subunits were in fact related in such FIG. 6. Schematic drawings showing approximately to scale a helical manner, we would be looking down the molecular the major features of the proposed mode of repressor-operator in- helix axis in the electron micrograph. The projection of such teraction. Note that each subunit would have two different sur- a helix would be round rather than ellipsoidal as we observe. faces, A and B, which can interact with the DNA, resulting thus Therefore, this model is not consistent with our present data in the possibility of two operator-binding sites (I and II) per re- pressor. The minor and major grooves of the DNA are sche- either. matically indicated as well as the size of operator protected by We would like to suggest another model for lac repressor- repressor from DNase digestion. Standard symbols are used to operator interaction that is consistent with the repressor's indicate the molecular 2-fold axes. The coincidence of the molecu- asymmetric shape, the 2-fold symmetry observed in the opera- lar 2-fold axes of the operator and the repressor binding site is tor sequence obtained by Gilbert and coworkers (13), and seen most clearly in a. The view shown in c corresponds.to the commonly observed features of subunit-subunit and protein- projected outline observed in the electron micrographs (Fig. 2). substrate interaction. We propose that the repressor binds to the operator with its long axis aligned with the long axis of the clear projected outline of the molecule observed in the the DNA (Fig. 5). Further, the DNA-binding site should be micrographs. between the four subunits and perhaps consist of a groove That the repressor molecule is asymmetric and 130- to or depression, as is commonly observed for hydrolytic en- 140-A long is suggested by the crystal packing and by calcu- zymes. That is, the repressor-operator complex might bear lation of the third dimension from the two dimensions ob- some similarities to a hotdog in a hotdog bun (Fig. 6a). Such tained from the micrograph. While a measurement of the size a DNA-binding site on the repressor would have the following of the white dumbbells in Fig. 2 would yield an underestimate consequences: (1) the binding site would have 2-fold sym- of the molecular dimensions due to the effects of the negative metry, (2) it could be as long as 130-140 A, (3) all four protein stain, the projected molecular outline can be expanded to subunits could simultaneously interact with the operator, and most nearly fill the unit cell. Measurement of the size of such (4) each subunit would have two different DNA-binding sur- tightly packed dumbbells indicates that two of the molecule's faces. The model would suggest the existence of two operator- dimensions are about 45 A and 60 A. Further, assuming the binding sites per tetramer if the repressor maintains perfect molecule to be a prolate ellipsoid of 150,000 molecular weight 222 symmetry. As with previous models, all four subunits and partial specific volume of 0.74 g/cm3, the third dimension are proposed to contribute to the specific and tight binding to can be directly calculated to be 135-140 A. operator, but unlike the previous models, unusual features of protein-protein or protein-substrate interaction are not sug- Quaternary structure of the repressor gested. The elongated shape and dumbbell-like projected profile that Our proposal that the binding site for operator show 2-fold we have observed for the repressor limits the possible quatern- symmetry would require that the operator DNA itself should ary structures it might have, especially if we assume that possess 2-fold symmetry and that the operator bind with its the repressor subunits are related by point group or quasi- 2-fold axis coincident with that of the repressor. The genetic point group symmetry. This is a reasonable assumption, since studies of Sadler and Smith suggested (17) and the sequence for all globular oligomeric proteins whose structures are known work of Gilbert and coworkers shows (13) that the operator the subunits are related by point group or quasi-point group contains sequences related by a 2-fold axis, in agreement with symmetry (16). For a tetramer there are only two possible our model for repressor (Fig. 6). That one operator constitu- point groups, D2 (three perpendicular molecular 2-fold axes) tive mutant lies in a region not related by a 2-fold axis (13) Downloaded by guest on September 28, 2021 Proc. Nat. Acad. Sci. USA 71 (1974) Structural Studies of lac Repressor Crystals 597

implies that the repressor possesses quasi-222 symmetry 1. Jacob, F. & Monod, J. (1961) J. Mol. Biol. 3, 318-356. rather than exact 222 symmetry, a situation that has been 2. Gilbert, W. & Muller-Hill, B. (1966) Proc. Nat. Acad. Sci. USA 56, 1891-1898. observed with some other proteins (18, 19). 3. Gilbert, W. & Muller-Hill, B. (1967) Proc. Nat. Acad. Sci. This model is not incompatible with the length of the opera- USA 58, 2415-2421. tor fragment protected from DNase digestion by repressor or 4. Riggs, A. D., Bourgeois, S., Newby, R. F. & Cohn, M. the genetic (9) and chemical (21) experiments implicating (1968). J. Mol. Biol. 34, 365-368. 5. Muller-Hill, B. (1971) Angew. Chem. Int. Ed. Engl. 10, 160- the amino-terminal region in operator binding. Gilbert (13) 172. finds that a piece of DNA 27 base-pairs long (about 90 A) is 6. Beckwith, J. R. & Zipser, D. (eds.) (1970) The Lactose protected by repressor from DNase digestion, which might Operon (Cold Spring Harbor Laboratory, Cold Spring Har- appear inconsistent with a repressor molecule 130- to 140-i bor, N. Y.). 7. Muller-Hill, B., Beyreuther, K. & Gilbert, W. (1971) in long. However, the binding site for operator could be longer Methods in Enzymology, eds. Grossman, L. & Moldave, K., (or shorter) than the piece of DNA protected from DNase if eds. in chief, Colowick, S. P. & Kaplan, N. 0. (Academic the nuclease binds and cleaves on the opposite side of the DNA Press, New York) Vol. XXI, Part D, pp. 483-487. from where the repressor binds. Further, the length of the 8. Riggs, A. D., Bourgeois, S. & Cohn, M. (1970) J. Mol. Biol. operator-binding site could be considerably shorter than the 53,401-417. 9. Adler, K., Beyreuther, K., Fanning, E., Geisler, N., length of the molecule, as is commonly observed for the poly- Gronenborn, B., Klemm, A., Muller-Hill, B., Pfahl, M. & peptide or polynucleotide binding sites of the hydrolytic en- Schmitz, A. (1972) Nature 237, 322-327. zymes. Genetic and chemical data (9, 21) have been used to 10. Crick, F. C. (1971) Nature 234, 25-27. argue that the first 50 residues or so from the amino terminus 11. Gierer, A. (1966) Nature 212, 1480-1481. 12. Sobell, H. M. (1972) Proc. Nat. Acad. Sci. USA 69, 2483- are involved in operator binding. These data do not really 2487. exclude the possibility that residues remote from the amino 13. Gilbert, W. & Maxam, A. (1973) Proc. Nat. Acad. Sci. USA terminus are also involved in DNA binding or even the possi- 70, 3581-3585. bility that the amino-terminal region is not directly involved 14. Platt, T. (1972) in Experiments in , ed. in DNA binding but in the allosteric transition between the Miller J. (Cold Spring Harbor Laboratory, Cold Spring Harbor, N. Y.), pp. 385-393. DNA-binding and non-DNA-binding repressor conformations. 15. Matthews, B. W. (1968) J. Mol. Biol. 33, 491-497. The model presented here does not require or exclude direct 16. Matthews, B. W. & Bernhard, S. A. (1973) "Structure and involvement of the amino-terminal residues in DNA binding Symmetry of Oligomeric " Annual Review of Bio- but would suggest that some other portions of the repressor physics and Bioengineering, eds. Mullins, L. J., Hagins, W. A. & Stryer, L. (Annual Reviews Inc., Palo Alto Calif.), might also be involved. Vol. 2. 17. Sadler, J. R. & Smith, T. F. (1971) J. Mol. Biol. 62, 139-169. 18. Steitz, T. A. (1971) J. Mol. Biol. 61, 695-700. We thank T. Platt and K. Weber for encouragement, helpful 19. Steitz, T. A., Fletterick, R. J. & Hwang, K. J. (1973) J. discussions, and repressor protein. We also thank W. Gilbert Mol. Biol. 78, 551-561. for discussing his operator sequence data with us. This work was 20. Labaw, L. W. & Davies, D. R. (1972) J. Ultrastruct. Res. supported by a grant from the Jane Coffin Childs Memorial 40, 349-365. Fund for Medical Research and by a U.S. Public Health Service 21. Platt, T., Weber, K., Ganem, D. & Miller, H. J. (1972) Grant GM-18268. Proc. Nat. Acad. Sci. USA 69, 897-901. Downloaded by guest on September 28, 2021