A Switch Between Two-, Three-, and Four-Stranded Coiled Coils in GCN4 Leucine Zipper Mutants
Pehr B. Harbury; Tao Zhang; Peter S. Kim; Tom Alber
Science, New Series, Vol. 262, No. 5138. (Nov. 26, 1993), pp. 1401-1407.
Stable URL: http://links.jstor.org/sici?sici=0036-8075%2819931126%293%3A262%3A5138%3C1401%3AASBTTA%3E2.0.CO%3B2-H
Science is currently published by American Association for the Advancement of Science.
Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/about/terms.html. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.
Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/journals/aaas.html.
Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.
The JSTOR Archive is a trusted digital repository providing for long-term preservation and access to leading academic journals and scholarly literature from around the world. The Archive is supported by libraries, scholarly societies, publishers, and foundations. It is an initiative of JSTOR, a not-for-profit organization with a mission to help the scholarly community take advantage of advances in technology. For more information regarding JSTOR, please contact [email protected].
http://www.jstor.org Sun Sep 9 23:07:30 2007 fit!@ RESEARCH ARTICLES
surements, each peptide was >90 percent A Switch Between Two-, Three-, helical at 4"C, neutral pH, and a concen- tration of 150 p,M (15), and each exhibited a cooperative thermal unfolding transition and Four-Stranded Coiled Coils in (Table 1). The midpoint of the thermal transition (T,) for each variant exceeded GCN4 Leucine Zipper Mutants that of the parental GCN4-pl peptide. Equilibrium analytical ultracentrifuga- Pehr B. Harbury, Tao Zhang,* Peter S. Kim, Tom Alber* tion (16) indicated that the peptides fall into three molecular weight classes: the Coiled-coil sequences in proteins consist of heptad repeats containing two characteristic peptides p-IL, p-11, and p-LI sedimented as hydrophobic positions. The role of these buried hydrophobic residues in determining the dimeric, trimeric, and tetrameric species, structuresof coiled coils was investigatedby studying mutantsof the GCN4 leucinezipper. respectively (Table 1). The oligomerization When sets of buried residues were altered, two-, three-, and four-helix structures were states of these peptides (p-IL, p-11, and formed. The x-ray crystal structure of the tetramer revealeda parallel,four-stranded coiled p-LI) were independent of peptide concen- coil. In the tetramer conformation, the local packing geometry of the two hydrophobic tration from 20 to 200 p,M. A derivative of positions in the heptad repeat is reversed relative to that in the dimer. These studies each peptide that contained the added se- demonstrate that conserved, buried residues in the GCN4 leucine zipper direct dimer quence Cys-Gly-Gly at the NH2-terminus formation. In contrast to proposals that the pattern of hydrophobic and polar amino acids also was synthesized (12). The cysteine in a protein sequence is sufficientto determinethree-dimensional structure,the shapes of residue permits disulfide bond formation, buried side chains in coiled coils are essential determinants of the global fold. and the two glycine residues provide a flexible linker. Pairing peptide monomers with a covalent disulfide bond did not change the oligomerization state of p-IL Recent evidence has suggested that the The structural variety of the coiled-coil and p-LI. In contrast, disulfide-bonded p-I1 three-dimensional structure of a protein is family and the functional requirements of sedimented with the molecular mass of a determined largely by the pattern of hydro- leucine zipper sequences suggest that geo- hexamer, consistent with the assignment of phobic (H) and polar (P) residues in the metric properties of buried, apolar amino p-I1 as a trimer in the absence of a disulfide amino acid sequence and is independent of acids may influence the overall structure of linkage. the geometric properties of the amino acid coiled coils. To investigate this possibility The peptides p-VI, p-VL, p-LV, and side chains that make up the pattern (1-3). we altered the hydrophobic core of a leu- p-LL populated multiple oligomerization This simplifying hypothesis, however, fails cine zipper molecule in a concerted fashion states (Table 1). The sizes of the complexes to account for a group of proteins composed and characterized the structures of the de- formed by these variants were determined of interacting, amphipathic a helices, the rivatives. by gel filtration using the peptides p-IL, coiled-coil family. Coiled-coil proteins Hydrophobic core mutants form two-, p-11, p-LI, and GCN4-pl as size standards have a characteristic seven-residue repeat, three-, and four-stranded structures. We (17). The peptides p-VI and p-VL ex- (a.b.c.d.e.f.g),, with hydrophobic residues systematically mutagenized the hydrophobic hibited concentration-dependent retention at positions a and d and polar residues core of the dimeric leucine zipper peptide times between 5 and 50 p,M, and deriva- generally elsewhere. Despite this shared HP GCN4-pl (12). With rare exceptions, hy- pattern, coiled-coil sequences adopt dimer- drophobic residues occupy the a and d posi- ic (4, 5), trimeric (6-8), and anti-parallel tions of the GCN4-p1 sequence, and polar tetrameric (9) conformations. residues appear at b, c, e, f, and g, generating In addition, parallel, dimeric coiled coils the characteristic (H.P.P.H.P.P.P), pat- exhibit strong preferences for specific amino tern. Because supercoiled a helices haxe acids at the hydrophobic a and d positions approximately 3.5 residues per turn, the of the heptad repeat. A striking example is spacing of the a and d positions three and provided by the leucine zipper motif, which four residues apart places .the H residues on functions to dimerize bZIP transcription one side of the helix. In the GcN4-~1 factors. Unlike other coiled coils, leucine dimer, the hydrophobic faces of two helides zippers contain leucine at -80 percent of pack against each other in a parallel orien- all d positions (5, 10). Multiple substitution tation (12, 13). Thus, the five amino acids of these leucines with similarly-sized hydro- at position a and the four leucines at position phobic residues often interferes with dimer- d from each monomer of GCN4-p1 form the ization function (11). apolar interface of the dimer (Fig. 1). We simultaneously changed four a P. B. Harbury is in the Department of Biological residues (Va19,Asn'6,Va123,Va130)and four Fig. 1. Helical wheel projection of residues Chemistry and Molecular Pharmacology, Haward Medical School, Boston, MA 021 15 and the Howard d residues (Le~~,Leu'~,Leu'~,Leu~~)of Met2 to GIu~~of the GCN4-pi sequence. View Hughes Medical Institute, Whitehead Institute, Depart- GCN4-pl to leucine, valine, or isoleucine is from the NH,-terminus, and residues in the ment of Biology, Massachusetts Institute of Technolo- (Met2 at the first a position was not first two helical turns are boxed or circled. 9 02142. gy, Cambridge Center, Cambridge, MA T. changed, Fig. 1) (14). These peptides were Heptad positions are labeled a through g. In the Zhang and T. Alber were In the Department of Bio- mutant peptides described here, the residues chemist~.Unlversitv of Utah School of Medicine. Salt named bv a two-letter code, the first letter in the dashed box at position a were collectively Lake ~itjl, UT 841i2. P. S. Kim is in the ~oward indicating the residueat the'four a positions Hughes Medical Institute, Whitehead Institute, Depart- changed to I, V, or L, and, separately, the ment of Biology, Massachusetts Institute of Technolo- and the second letter indicating the residue residues in the dashed box at position d were gy,-. 9 Cambridge Center, Cambridge, MA 02142. at the four d positions. The sequences were changed to I, V, or L (41). For example, p-LI *Present address: Department of Molecular and Cell GCN4-~1, contains leucine at the four boxed a positions Biology, University of California, Berkeley, CA 94720 On the basis of circular dichroism mea- and isoleucine at the four boxed d positions.
SCIENCE VOL. 262 26 NOVEMBER 1993 tives of p-VI, p-LV, and p-LL that con- Unlike two- and three-stranded coiled peptide monomers and the disulfide bonds tained an NH2-terminalCys-Gly-Gly disul- coils, all well-characterized examples of to exchange (Fig. 2B) (21). Only ho- fide linkage eluted as multiple species. four interacting a helices exist in an anti- modimers were observed at equilibrium. Helix orientation. The observation that parallel four-helix bundle conformation Assuming that the glycyl linkers allow the the p-IL peptide remains dimeric with an (Fig. 2A) (20). To determine the helix terminal cysteines to assort randomly, the NH2-terminal disulfide linkage (Table 1) orientation of the p-LI tetramer, variants of results indicate that the p-LI peptide assem- indicates that the helices are parallel. Con- the p-LI peptide with the sequence Cys- bles into a conformation containing four sequently, the conformation of p-IL likely Gly-Gly at the NH2-terminus (denoted parallel a helices (22). To investigate the resembles the structure of GCN4-pl. The p-LI-N) or with the sequence Gly-Gly-Cys basis for the switch between parallel tetra- helices of the trimeric peptide p-I1 also are at the COOH-terminus (denoted p-LI-C) mer, dimer, and trimer conformations, the parallel because a two-dimensional double were synthesized. The disulfide-bonded x-ray crystal structure of the p-LI peptide quantum-filteredcorrelation spectrum (DQF p-LI-N-p-LI-C heterodimer was purified was determined at 2.1 A resolution (Table COSY) of the peptide showed only one and placed in redox buffer to allow the 2) (23). magnetic environmentfor each residue (18), and the p-I1 peptide crystallized on a three- fold rotation axis (19). Thus, p-I1 assumes the same oligomeric structure as the trimeric stalk of influenza hemagglutinin (8).
A Anti-parallelfour-helix bundle
Flg. 3. The pLI peptide All parallel helices forms a parallel, fwr- stranded coiled coil. (A) A portion of the 2F,-F, elec- tron dens* map showing the d level containing Ilel@. View is from the NH2-termi- nus down the superhelix axis. (8) An axial view of the p-U tetramer next to the GCNepl dirner. The van der Waals (VDW) sur- faces of side chains at e (purple) and d (green) are depicted. (C) Side view of the pU tetramer showing the VOW surfaces of resi- dues at a (purple) and d (green) superimposed on the helix back- bone. The NH,-terminal methiu Flg. 2. The p-LI peptide forms a parallel tetra- nine layer is yellow. (D) Helical mer in solution. (A) Given random pairing of wheel representation of residues terminal cysteine residues, the anti-parallel 2 to 32 of the p-LI tetrarner. Vi four-helix bundle conformation should produce is from the NH2-terminus,and res- a mixture of N-ss-C, N-ss-N, and C-ss-C disul- idues in the first two heiical turns fide bonds. The parallel conformation should are boxed (Met2) or circled. Hep produce only N-ss-N and C-ss-C disulfide tad positions are labeled a bonds. (6)The disulfide-bonded p-LI-N-p-LI-C through g. ll-te 4, o,, and + heterodimer (N-ss-C) rearranges to form ho- variables from a coiled-coil pa- modimers (N-ss-N and C-ss-C).The disulfide- rameterization suggested by bonded heterodimer of p-LI-N (a variant of p-LI Crick are illustrated (Table 2) with the sequence CGG added to the NH,- (26). terminus) and p-LI-C (a variant of p-LI with the sequence GGC added to the COOH-teriminus) was purified and incubated in redox buffer to allow the exchange of disulfide bonds (21).At the indicated times a portion of the sample was removed, quenchedwith acid,and analyzed by HPLC. The p-LI-C peptide has a larger extinc- tion coefficientthan the pLI-N peptide because it contains an additional tyrosine residue. K
SCIENCE VOL. 262 26 NOVEMBER 1993 Structure of the tetramer. The p-LI The gross differences between the struc- (13). The tetramer also exhibits several tetramer consists of four ~arallela helices tures of the p-LI tetramer and the GCN4- types of ion pairs not present in the GCN4- wrapped in a left-handed superhelix (Fig. pl dimer (1 3) may be summarized as fol- pl dimer. These include four of eight pos- 3). The helices create a cylinder that is lows: (i) Compared to the helices of the sible interhelical g to b salt bridges (for -27 A wide and -48 A long. An approx- GCN4-pl dimer, diagonally related helices example, LysD8-G1~A10,Fig. 4D) and five imate fourfold axis of symmetry coincides of the p-LI tetramer have the same relative of eight possible interhelical c to e salt with the superhelical axis (24). orientation but are 5.5 A further apart (Fig. bridges (for example, HisB18-G1~CZ0,Fig. The leucine and isoleucine side chains 3B). (ii) Adjacent helices of the p-LI tet- 4E). Finally, a charge-stabilized hydrogen at the a and d ~ositions~ointinto the ramer are separated by approximately the bond forms between ArgBz5and the main center of the tetramer. Cross-sectional lay- same distance as the GCN4-pl dimer heli- chain carbonyl oxygen of LeuCz3. ers containing- leucine at the a ~ositions ces, but each helix is rotated by -45" More surface area is buried in the tetra- alternate with layers containing isoleucine around its own axis toward the center of the mer (1640 A2per helix) than in the dimer at the d positions (Fig. 3C). The dihedral tetramer (Fig. 3B). Helical parameters are (900 AZper helix) (28). Compared to the angles X, and xz of all leucines at a and compared in Table 3 (26). side chains of isolated helices, residues at all isoleucines at d are approximately Helix interactions. On the basis of dis- the a, d, e, and g positions of the tetramer (-60,180), corresponding to the most pop- tance criteria (27), residues on the surface are substantially buried (>66 percent) ; res- ulated rotamer for both amino acids (25). of the tetramer appear to form interhelical idues at the b and c positions are partly The side chains of MetZ form the most ion pairs. Of 12 possible interhelical salt buried (-15 percent); and the f positions NHz-terminal a layer. bridges between residues at the e position of remain completely exposed. The e and g A cavity in the middle of each leucine one heptad and the g position of the pre- residues in the tetramer are almost as buried
and isoleucine layer forms a continuous ceding heptad [for example, G1uB6-ArgA', as the a and d residues of the dimer (28).~, central channel. The radius of the channel (27)], five are seen in the tetramer crystal Similarly, the b and c residues of the tetra- varies from 1.0 to 1.3 A and therefore structure. A similar frequency of interheli- mer are almost as buried as the e and -rr excludes a 1.4 A radius water-molecule cal e to g interactions was found in the residues of the dimer. probe. No electron density was seen in the crystal structure of GCN4p1, which con- The tetramer interface shows knobs-in- channel. tains three of six possible e to g salt bridges to-holes packing between helices (Fig. 4). As proposed by Crick (29), knobs formed by the side chains of one helix fit into holes Table 1. Core mutants of GCN4-pl form stable two-, three-, and four-helix structures. formed by the spaces between side chains on the neighboring helix. Looking from the Positions* No. of helicest -[f'l222 Tm TmGdmC1t NHz-terminus down the superhelix axis, (deg cm2 dmol-l) ("c) PC) a d Unmodified SS each leucine knob at an a position packs into a hole formed by the g and,a residues of GCN4-p1 33,300 53 <0 2 2 the counterclockwise-related monomer I L 32,400 >I00 77 2 2 (Fig. 4D) and by two d residues in adjacent I I 32,400 >I00 70 3 6 layers along the superhelix axis. Similarly, L I 30,600 >I00 94 4 4 each isoleucine knob at a d position packs V I 22,5000 73
SCIENCE VOL. 262 26 NOVEMBER 1993 1403 the dimer and the a levels of the tetramer A
4C). A third class of knobs-into-holes ii- teraction appears at the a and d positions of parallel trimeric coiled coils (8). At both levels of the influenza hemagglutinin tri- mer, the Ca-CP bond of each knob makes a -60' angle with the Ca-Ca vector at the base of the corresponding hole. We define this arrangement as acute knobs-into-holes packing (Fig. 4A). Because the peptides p-IL, p-11, and p-LI differonly by volume-conserving hydropho- bic substitutions at the buried a and d positions, packing interactions at a and d must mediate the switch between the di- mer, trimer, and tetramer conformations. The geometric relationship between the dimer and tetramer structures suggests an explanation for the different oligomeriza- 1 tion of the peptides p-IL and p-LI. The inversion of sequence at the a and d posi- tions (Ile, Leu to Leu, Ile) coincides with an inversion of the packing geometry at a and d (parallel-perpendicularto perpendic- ular-parallel) in the two- and four-stranded conformations. Evidently, parallel packing has a geometric preference for isoleucine, or perpendicular packing has a geometric pref- erence for leucine, or both (30). Recent measurements of amino acid preferences in a dimeric coiled-coil model indicate that both types of packing exhibit the expected bias. Isoleucine is strongly favored (-0.4 kcal mol-') over leucine at Fig. 4. Three types of knobs-into-holes packing and electrostatic interactions. (A) Schematic the parallel geometry and leucine is more drawing (not to scale) showing the relative positions of the Ca-Cp and Ca-Ca vectors for weakly favored (-0.1 kcal mol-') over perpendicular, parallel, and acute knobs-into-holes packing. (B) Superposition of perpendicular isoleucine at the perpendicular geometry packing in a GCN4-pl dimer d level (green) and a p-LI tetramer a level (white). The Ca-Cp bond (31). If the a position of the tetramer is of each knob (thck red line) makes a -90" angle with the Ca-Ca vector at the base of the hole on assumed to have the same residue specificity the neighboring helix (thick yellow line). The two dimer d positions overlay on an a position and a as the geometrically similar d position of g position of the tetramer (white VDW surfaces). (C) Superposition of parallel packing in a GCN4-pl the dimer (and similarly position d of the dimer a level (green) and a p-LI tetramer d level (white). The Ca-Cp bond of each knob (thick red tetramer is assumed to have the same resi- line) is parallel to the Ca-Cavector at the base of the hole on the neighboring helix (thick yellow line). The two dimer a positions overlay on a d position and an e position of the tetramer (white VDW due specificity as position a of the dimer), surfaces). (D) g to b salt bridges in the tetramer a level (Leug). Lysines at g positions (blue VDW then it follows that the two- to four-strand- surfaces) make salt bridges with glutamates at b positions (red VDW surfaces). White VDW surfaces ed transition is driven primarily by prefer- idenfity leucines at a positions. (E) c to e salt bridges in the tetramer d level (llei9). Glutamates at ence for isoleucine at the levels with paral- e positions (red VDW surfaces) make salt bridges with histidines at c positions (blue VDW surfaces). lel packing. White VDW surfaces identify isoleucines at d positions. However, at least two arguments support the opposite conclusion, that discrimi- nation against isoleucine at perpendicular would explain why the p-I1 peptide forms a GCN4-pl peptide forms a dimer, but sub- positions dominates the conformational trimer. The dimer and tetramer conforma- stitution of Asn16 (at an a position) with switch. First, dimeric, fibrous coiled-coil tions would each lace four isoleucines of valine causes the resulting peptide p-VL to sequences show a strong bias against isoleu- p-I1 in a perpendicular geometry, while the populate both dimeric and trimeric confor- cine residues at d (I)positions and show trimer conformation places all eight isoleu- mations (Table 1). Despite this heteroge- no preference for isoleucine or leucine at a cines in the acute geometry. Sequences of neity, p-VL exhibits a significantly higher (11) positions (32). Second, modeling stud- trimeric coiled coils show no strong bias for apparent T, than GCN4-pl. Thus, Asn16 ies indicate that isoleucine must ado~tan leucine or isoleucine at either the a or d imposes specificity for the dimer structure at infrequently observed rotamer (- ,-) to positions (6). the expense of stability (35). occupy the perpendicular position of a di- In addition to efficient packing, a buried Residues at the b, c, e, and g positions meric coiled coil, while the most common hydrogen bond between Asn16 residues spe- are more buried in the tetramer structure rotamer of leucine may be accommodated cifically favors the dimer conformation of than in the dimer. Thus, the amino acid at the parallel position (33). Poor packing GCNCpl, the parent molecule of the sequence at these positions can also influ- of isoleucine at perpendicular positions also coiled coils described here (1 3, 34). The ence oligomerization state (31). SCIENCE VOL. 262 26 NOVEMBER 1993 Table 3. Superhelical parameters calculated from the refined dimer (13) and tetramer structures son, W. A. Baase, F. W. Dahlquist, B. W. Mat- using a parameterization suggested by Crick (26).The Ro,w,, and I$ variables are illustrated in Fig. thews, Biochemistry30, 11521 (1991), D. Shortle, 3D. W. E. Stites, A. K. Meeker, ibid. 29, 8033 (1990); W. E. Stites,A. G. Gittis, E. E. Lattman, D. Shortle, J. Mol. Biol. 221, 7 (1991). Parameter Dimer Tetramer 2. J. U. Bowie, R. Luthy, D. Eisenberg, Science 253, 164 (1991). Supercoil radius, Ro (A) 3. K. F. Lau and K. A. Dill, Macromolecules 22, 3986 Amino acids per supercoil turn, o, (1989); H. S. Chan and K.A. Dill, Proc. Natl.Acad. Supercoil pitch (A) Sci. U.S.A. 87, 6388 (1990); A. Sikorski and J. Helix crossing angle Skolnick, Biopolymers 28, 1097 (1989). Radius of curvature 4. J. F. Conway and D. A. D. Parry, Int. J. Biol. (A) Macromol. 12, 328 (1990). Position a orientation angle, I$ 5. J. C. Hu and R. T. Sauer, in Progress in Nucleic Acids and Molecular Biology, X. X. Eckstein and X. X. Lilley, Eds. (Springer-Verlag, Berlin, 1992), vol. 6, pp. 82. Protein structure. Recent studies have 1). In the case of the trimer conformation, 6. J. F. Conway and D. A. D. Parry, Int. J. Biol. suggested that the tertiary folds of globular p-11, though less stable than p-LL, exhibits Macromol. 13, 14 (1991). 7. P. K. Sorger and H. C. Nelson, Cell 59, 807 proteins are insensitive to the details of higher specificity for the trimer fold. The (1989); R. Peteranderl and H. C. Nelson, Bio- packing, and are determined instead by results suggest that by ignoring specific chemistry 31, 12272 (1992), B. Lovejoy et al., hydrophobic-polar (HP) and secondary packing interactions, HP-2s design meth- Science 259, 1288 (1993). 8. 1. A. Wilson, J. J. Skehel, D. C. Wiley, Nature 289, structlre (2s) patterns in the amino acid ods fail to produce high conformational 366 (1981);W. I. Weis, A. T. Brunger, J. J. Skehel, sequence. Several globular proteins, for ex- selectivitv and sometimes oredict seauences D. C. Wiley, J. Mol. Biol. 212, 737 (1990). ample, accommodate multiple hydrophobic with less 'than optimal stability. 9. D. W. Banner, M. Kokkinidis, D. Tsernoglou, ibid. 196, 657 (1987). substitutions at core positions without Practical implications. The mutants of 10. W. H. Landshultz, P. F. Johnson, S. L. McKnight, adopting new folds (1). In addition, HP-2s GCN4-pl clarify the functions of conserved Science 240, 1759 (1988). sequence patterns, when compared with an features of leucine zipper sequences. The leu- 11. J. C. Hu, E. K. O'Shea, P. S. Kim, R. T. Sauer, ibid. HP-2S template derived from a known cine repeat at d positions and the preponder- 250, 1400 (1990); T. Kouzarides and E. Ziff, Nature 336, 646 (1988); M. Neuberg, M. Schuer- three-dimensional structure, are more sen- ance of p-branched amino acids at a positions man, J. B. Hunter, R. Muller, Science 338, 589 sitive indicators of structural homology favor dimer formation due to packing consid- (1989); R. Gentz, F. Rauscher, C. Abate, T. Cur- than is direct sequence comparison (2). erations (Table I), while conserved aspar- ran, ibid., p. 1695; R.Turner and R. Tjian, ibid., p. 1689; L. J. Ransone, J. Visvader, C. P. Sassone, I. Finally, bipartite lattice models of proteins, agines at a positions direct dimerization by M. Verma, Genes Dev. 3, 770 (1989); M. Schuer- consisting of H-type and P-type residues, forming buried hydrogen bonds. man et a/., Cell 56, 507 (1989); V. J. Dwarki, M. Montminy, I. M. Verma, EMBO J. 9, 225 (1990). fold to single, compact conformations un- Together" with analvses of coiled-coil 12. E. K. O'Shea, R. Rutkowski, P. S. Kim, Science der potentials that maximize HH contacts sequences (4, 6), our results also suggest 243, 538 (1989) (3). Within the limitations of the comouter that predictions of coiled-coil oligomeriza- 13. E. K. O'Shea, J. D. Klemm, P. S. Kim, T. Alber, models, the drive to bury hydrophobic res- tion can be based in part on the distribution ibid. 254, 539 (1991) idues can specify unique tertiary structures. of P-branched residues at the a and d 14 Peptideswere synthesized and purified as in D. J. Lockhart and P. S. Kim, Science257,947 (1992). The heptad HP-2s pattern is clearly not positions. The occurrence of P-branched The identities of the peptides were confirmed by sufficient to guide a coiled-coil sequence to residues at d positions disfavors dimers, mass spectrometry (Finnigan MAT), and all mo- a single structure. Unlike globular proteins, while P-branched residues at a positions lecularweights were found to be within 1 dalton of the expected mass.The peptides have acetylated however, coiled coils have a very simple HP should disfavor tetramers. The presence of NH,-termlni and free COOH-termini pattern and contain unconnected units of B-branched residues at both the a and d 15. CD spectra were measured on an AVIV 60DS secondary structure. Such properties allow positions facilitates trimer formation. spectropolarimeter Measurementsof [0],,, were made at 4°C in 50 mM phosphate (pH 7.0), 150 coiled-coil monomers to assemble in multi- The tetramer structure. which contains mM NaCl, and 10 &M peptide. Thermal melts ple configurations. Thus, rearrangement of a large axial cavity, may serve as a soluble were performed in the same buffer and also with the hydrophobic core in response to muta- model for membrane ion channels. Related the addition of 3 M guanidinium chloride (GdmCI) to facilitate unfolding of the extremely stable tions, which globular proteins appear to sequences that contain serine at the a and d coded coils. Values of T,, were estimated from accomplish while conserving tertiary fold, positions have been shown to conduct small [0],,, versus temperature data (in two-degree can be manifested in coiled coils as dramat- cations (39). Substitutingserine for leucine steps) by evaluating the maximum of d[0]ldT-' and isoleucine residues in the crystallo- [C. R. Cantor and P. R. Schimmel, Biophysical ic changes in overall structure. Consistent Chemistry (W H Freeman and Company, New with this argument, conservative core mu- graphic coordinates of the p-LI tetramer York, 1980),vol. 3, pp. 11321. Peptide concentra- tations often result in loss of function for generates a model with an enlarged interior tions were determined spectrophotometrically [H. coiled coils (I but not for their globular channel lined by the hydroxyl groups of the Edelhoch, Blochemistry 6, 1948 (1967)l with an I) extinction coefficient of 1280 cm-'Mi at 280 nm. counterparts (1). serine side-chains. 16. Analyt~calultracentr~fugat~onmeasurementswere Recent attempts to design helical pro- Finally, the short, modular peptides made on a Beckman XLA centrifuge equipped teins based on HP-2S patterns have pro- p-11, p-LI, and p-IL may be used to design with an An-60Ti rotor Ten or fifteen data sets (20 to 200 yM at two or three rotor speeds) were fit duced molecules that lack unique packing two-, three-, and fourfold oligomers of at- simultaneously to a single molecular welght with (36). For example, the four-helix bundle a4 tached sequences. Addition of p-VL, alter- the program HID4000 [M. L. Johnson, J. J. Cor- exhibits a well-defined secondary structure natively, would produce a hybrid capable of reia, D. A Yphantls, H. R. Halvorson, Biophys. J. 36,575 (1981)], provided by M. L. Johnson and J. and is very stable, but the leucine side interconvertine- between dimer and trimer Lary. Molecular welghts were determined under chains in the core of the molecule have states. Such inherent structural heterogene- two different conditions. (i) 50 mM sodlum phos- fluctuating conformations (37). Within the ity could form the basis for a regulatory phate (pH 7.1), 500 mM NaCl at 4°C and (il) 50 mM sod~umphosphate (pH 2.0), 100 mM NaCl at GCN4-pl variant family, HP-2S design switch (40). 0°C. Samples were dialyzed against the appropri- criteria predict that the p-LL sequence ate buffer for at least 12 hours before each exper- should form an optimal coiled coil (38). For REFERENCES AND NOTES iment. Speclflc volumes and solvent density were the dimer and tetramer conformations, calculated accord~ngto T M. Laue, B D. Shah,T 1 W. A Lim and R. T Sauer, J. Mol. Biol 219, 359 M R~dgeway,S. L. Pelletler, in Analytical Ultra- however, p-IL and p-LI exhibit equal or (1991),J. H. Hurley,W. A. Baase, B.W. Matthews, centrifugation fn Biochemistry and Polymer Sci- higher thermal stabilities than p-LL (Table ibid. 224, 1143 (1992); S. Dao-pin, D. E. Ander- ence, S. Harding, A Rowe,J. Horton, Eds (Royal
SCIENCE VOL. 262 26 NOVEMBER 1993
Wisniewski, C. Wu, Science 259, 230 (1993). Staley, B. Tidor, and J. Weissman for discussions core promoter is scficient for specific initi- 41. Abbreviations for the amino acid residues are: A, and critical reading of the manuscript. PSK, is a Ala; C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, HIS; Pew Scholar in the Biomedical Sciences. This ation and for response to the two 'ysterns. .. I, Ile; K, LVS; L, ~eu;M, Met; N,Asn; P, pro: Q.Gln: research was su~~ortedby arants from the Na- known to regulate transcription under dif- R, Arg, S; Ser; T. Thr; V, Val; W, Trp; Y, Tyr. tional Institutes 0; ~ealth(~~24162to P.S.K.and ferent nutritional conditions-namely, 42. We thank R. Rutkowski and M. Burgess for pep- GM48958 to T.A.) and by grants from the Lucille tide synthesis, B. Santarsiero of Molecular Struc- p, ~~~k~~ charitable ~~~~t and the ~~~~i~~~ growth rate-de~endent and strin- tures Corporation for collecting high resolution Cancer Society (T.A.). gent control (13, 14)-the region upstream data from the p-LI crystals, and C. Carr, A. Coch- of the core promoter is largely responsible ran. D. Lockhart, Z. Y. Peng. B. Schulman, J. 22 June 1993; accepted 26 October 1993 for its high activity (4, 9, 13, 15, 16). A 20-base pair (bp) region rich in (A + T), to which we refer as the upstream (UP) element, is located immediately upstream A Third Recognition Element in of the core promoter (Fig. 1). The UP element increases rrnB PI activity by a Bacterial promoters: DNA Binding factor at least 30 in the absence protein factors other than RNAP (9, 17). ~ h ~ ~ ~ ~ l ~ ~ ~ ~ t i ~ ~ ~ ~ t ~ ~ t ~ d b ~ ~ ~ ~ ~ i ~ by the a Subunit of RNA Polymerase footprinting experiments, and replacement of the UP-element with non-rrn~DNA Wilma ROSS,Khoosheh K. Gosink, Julia Salomon, results in severe reduction of protection in Kazuhiko Igarashi, Chao ZOU,Akira Ishihama, the upstream region (8, 18, 19). Therefore, Konstantin Severinov, Richard L. Gourse* the core and UP element together can be considered an extended promoter (Fig. 1) A DNA sequence rich in (A + T), located upstream of the -10, -35 region of the (9). The region adjacent to the UP element Escherichia coli ribosomal RNA promoter rrnB PI and called the UP element, stimulates (between bp -60 and bp -150) contains transcription by a factor of 30 in vivo, as well as in vitro in the absence of protein factors binding sites for the activator protein Fis, other than RNA polymerase (RNAP). When fused to other promoters,such as lacUV5, the which results in increasing the activity of UP element also stimulates transcription, indicatingthat it is a separable promoter module. the promoter by a factor of 10 (8, 16, 20, Mutations in the carboxyl-terminalregion of the a subunit of RNAP prevent stimulation of 21). Fis is not required for stimulation of these promoters by the UP element although the mutant enzymes are effective in tran- transcription by the UP element (9). scribing the "core" promoters (those lacking the UP element). Protection of UP element The u70 subunit of RNAP holoenzyme DNA by the mutant RNAPs is severely reduced in footprintingexperiments, suggestingthat (a,PP1u) interacts with the -10 and -35 the selective decrease in transcription might result from defective interactions between a hexamers (22). However, the region or and the UP element. Purified a binds specifically to the UP element, confirmingthat a acts regions of RNAP required for UP element directly in promoter recognition. Transcription of three other promoters was also reduced recognition have not been defined. On the by the COOH-terminal a mutations.These results suggest that UP elements comprise a basis of studies with mutant derivatives of third promoter recognition region (in addition to the -10, -35 recognition hexamers,which the a subunit, it has been proposed that a interact with the u subunit) and may account for the presence of (A + T)-rich DNA interacts directly with certain transcription upstreamof many prokaryotic promoters.Since the same a mutations also block activation factors, leading to stimulation of promoter by some transcription factors, mechanisms of promoter stimulation by upstream DNA activity (23-26). We therefore used mu- elements and positive control by certain transcription factors may be related. tants of a to investigate the role of this subunit in UP element function although transcription activation in this case is achieved by a DNA element rather than by The strength of promoters recognized by factors other than RNAP (3-9), and w- a trans-acting protein. Two mutant forms of Eu70, the most abundant of the E. coli gions rich in (A + T) have been noted the 329 amino acid a subunit, COOH- RNAP holoenzymes, can be correlated to a upstream of many promoters (2, 10, 11). terminal truncations of 73 or 94 amino considerable extent with their similarity to The rrnB P1 promotci'i is representative acids (a-256 or a-235, respectively), are consensus recognition hexamers in the core of a class of seven rRNA promoters in E. stable in vivo and assemble into holoen- promoter region, centered approximately coli that together account for more than zyme (27). Furthermore, reconstituted 10 and 35 bp upstream of the start site of half of the transcription in the cell at high a-235 or a-256 core enzymes (a,PPf), transcription, and the spacing between growth rates (12). Although the rrnB P1 prepared in vitro from purified subunits, these hexamers (I ). Nevertheless, it has been proposed that sequences outside of the Fig, ,. The rrnB pro. UP Core core promoter region can modulate promot- meter, ~h~ extended pro- Fis sites element promoter er activity (2). Upstream sequences have meter region includes ele- 111 I1 I -35 -10 been shown to increase the activities of ments recognized by RNA -60 -40 several Escherichia coli or Bacillus subtilis polymerase: the core pro- H H I I promoters in vitro in the absence of protein meter, which consists of -150 -160 -50 +I -10 and -35 consensus W. Ross, K. K. Gosink, J. Salomon, and R L. Gourse hexamers (filled boxes), __-___------Extended promoter .- . are In the Department of Bacteriology, University of and the UP element, which Wisconsin-Madison, 1550 Linden Drive, Madison, WI UP Element -35 -10 Of bp -40 -60 I 53706. K. Igarashi, C.Zou, and A. lshihama are in the consists to -60 -50 -40 -30 -20 +? Department of Molecular Genetics, National Institute (shaded box) (9) Three ~ L ~ ~ ~ ~ ~ ~ ~ T T ~ T ~ T ~ ~ ~ ~ T ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ of Genetics, Mishima, Shizuoka 41 1, Japan K. Sever- binding sites for the tran- inov is at the Public Health Research' Institute, 455 scriptionalactivator protein Fis (open boxes) occur upstream of the UP element.Site I is responsible First Avenue, New York, NY 10016. for most of the activation by Fis at this promoter (20,21).The sequence of the extended promoter *To whom correspondence should be addressed is indicated below the diagram.
SCIENCE VOL. 262 26 NOVEMBER 1993 1407