suggest additional photochemical mecha-nisms by which OClO can contribute to 03depletion, should OClO be adsorbed on orsolvated within aerosols.

Work in progress (52) measures thesticking coefficients for OCIO on ice sur-faces to obtain the OCIO content of atmo-spheric aerosols. Recent photochemicalstudies of OCIO in warm (80 to 150 K) icematrices find quantitative conversion ofOCIO to ClOO (53). Possible atmosphericconsequences of OClO in aerosols includevertical redistribution of Cl in the Antarc-tic vortex, release of active Cl by reactionwith HCI, and heterogeneous photochemi-cal 03 depletion.

36. ,_ ibid., p. 163.37. J. B. Coon, ibid. 14, 665 (1946).38. S. Michielsen et al., ibid. 74, 3089 (1981).39. T. Baumert, J. L. Herek, A. H. Zewail, ibid. 99, 4430

(1993).40. E. Ruhi, A. Jefferson, V. Vaida, J. Phys. Chem. 94,

2990 (1990).41. E. Bishenden and D. J. Donaldson, J. Chem. Phys.

99, 3129 (1993).42. H. F. Davis and Y. T. Lee, J. Phys. Chem. 96, 5681

(1992).43. W. G. Lawrence, K. C. Clemitshaw, V. A. Apkarian,

J. Geophys. Res. 95,18591 (1990).44. J. H. Glownia, J. Misewich, P. P. Sorokin, in Super-

continuum Lasers, R. R. Alfano, Ed. (Springer-Ver-lag, Berlin, 1990), pp. 1-4.

45. A. J. Colussi, R. W. Redmond, J. C. Scaiano, J.Phys. Chem. 93, 4783 (1989).

46. A. J. Colussi, ibid. 94, 8922 (1990).

47.

48.49.

50.

51.52.

53.

54.

K. Johnsson, A. Engdahl, P. Ouis, B. Nelander, J.Mol. Chem. 293,137 (1993).

,- J. Phys. Chem. 96, 5778 (1992).C. Reichardt, Solvents and Solvent Effects in Organ-ic Chemistry (VCH, New York, 1988).R. L. Jones et al., J. Geophys. Res. 94, 11529(1989).J. G. Anderson et al., ibid., p. 1 1480.C. J. Purcell, J. Conyers, P. Alapat, R. Parveen, J.Phys. Chem., in press.J. T. Roberts, V. Vaida, L. A. Brown, J. D. Graham, inpreparation.This work was supported by grants from the NationalScience Foundation (V.V., ATM-9222651, andJ.D.S., CHE-9306485). We thank our co-workers onthis project; R. C. Dunn, B. Flanders, C. Foote, J. L.Anderson, E. Richard, E. Ruhl, A. Jefferson, G. Frost,L. Brown, and S. Solomon. We also thank K. Peter-son for many helpful discussions.

REFERENCES AND NOTES

1. J. C. Farman, B. G. Gardiner, J. D. Shanklin, Nature315, 207 (1985).

2. A. F. Tuck, T. R. Watson, E. P. Condon, J. J. Margi-tan, 0. B. Toon, J. Geophys. Res. 94,11181 (1989).

3. S. Solomon, Nature 347, 347 (1990).4. World Meteorlogical Organization Report 20 (1990),

vol. 1.5. J. G. Anderson, W. H. Brune, M. H. Proffitt, J. Geo-

phys. Res. 94, 11465 (1989).6. W. H. Brune, J. G. Anderson, K. R. Chan, ibid., p.

16649.7. S. Solomon, R. W. Sanders, M. A. Carroll, A. L.

Schmeltekopf, ibid., p. 11393.8. C. Schiller et al., Geophys. Res. Lett. 17, 501 (1990).9. G. Brasseur and S. Solomon, Aeronomy of the

Middle Atmosphere (Reidel, Dordrecht, Holland,1986).

10. S. Solomon, R. R. Garcia, F. S. Rowland, D. J.Wuebbles, Nature 321, 755 (1986).

11. M. A. Tolbert, Science 264, 527 (1994).12. C. R. Webster et al., ibid. 261, 1130 (1993).13. D. R. Hanson and A. R. Ravishankara, J. Geophys.

Res. 96, 5081 (1991).14. M. J. Molina, T.-L. Tso, L. T. Molina, F. C.-Y. Wang,

Science 238, 1253 (1987).15. L. T. Molina and M. J. Molina, J. Phys. Chem. 91,

433 (1987).16. S. P. Sander, R. R. Friedi, Y. L. Yung, Science 245,

1095 (1989).17. M. B. McElroy, R. J. Salawitch, S. C. Wofsy, J. A.

Logan, Nature 321, 759 (1986).18. A. A. Turnipseed, J. W. Birks, J. G. Calvert, J. Phys.

Chem. 95, 4356 (1991).19. H. S. Muller and H. Willner, Inorg. Chem. 31, 2527

(1992).20. V. Vaida, S. Solomon, E. C. Richard, E. Ruhl, A.

Jefferson, Nature 342, 405 (1989).21. K. A. Peterson and H.-J. Werner, J. Chem. Phys. 96,

8948 (1992).22. S. Baer et al., ibid. 95, 6463 (1991).23. A. Arkel and 1. Schwager, J. Am. Chem. Soc. 89,

5999 (1967).24. F. J. Adrian, J. Bohandy, B. F. Kim, J. Chem. Phys.

85, 2692 (1986).25. R. C. Dunn, B. N. Flanders, V. Vaida, J. D. Simon,

Spectrochim. Acta A 48, 1293 (1992).26. V. Vaida, K. Goudjil, J. D. Simon, B. N. Flanders, J.

Mol. Liq. 61, 133 (1994).27. R. C. Dunn and J. D. Simon, J. Phys. Chem., in press.28. _ , J. Am. Chem. Soc. 114, 4856 (1992).29. R. C. Dunn, J. Anderson, C. S. Foote, J. D. Simon,

ibid. 115, 5307 (1993).30. A. W. Richardson, R. W. Redding, J. C. D. Brand, J.

Mol. Spectrosc. 29, 93 (1973).31. K. Miyazaki, M. Tanoura, K. Tanaka, T. Tanaka, ibid.

116, 435 (1986).32. C. Paillard, G. Dupre, H. AlAiteh, S. Youssefi-Stitou,

Ann. Phys. (Paris) 14, 641 (1989).33. J. L. Gole, J. Phys. Chem. 84,1333 (1980).34. J. A. Jafri, B. H. Lengsfield Ill, C. W. Bauschlicher Jr.,

D. H. Phillips, J. Chem. Phys. 83,1693 (1985).35. E. C. Richard and V. Vaida, ibid. 94, 153 (1991).

RESEARCH ARTICLEt

Bent Helix Formation BetweenRNA Hairpins with

Complementary Loops

John P. Marino, Razmic S. Gregorian Jr., Gyorgyi Csankovszki,Donald M. Crothers*

The initial interaction between the ColEl plasmid specific transcripts RNA and RNA II,

which function as antisense regulators of plasmid replication, comprises a transientcomplex between complementary loops found within the RNA secondary structures.Multidimensional heteronuclear magnetic resonance spectroscopy was used to charac-terize complexes formed between model RNA hairpins having seven nucleotide com-plementary loops. Seven base pairs are formed in the loop-loop helix, with continuoushelical stacking of the loop residues on the 3' side of their helical stems. A sharp bendin the loop-loop helix, documented by gel electrophoresis, narrows the major groove andallows bridging of the phosphodiester backbones across the major groove in order toclose the hairpin loops at their 5'-ends. The bend is further enhanced by the binding ofRom, a ColEl encoded protein that regulates replication.

Regulation of the replication of the Esch-erichia coli plasmid CoIEl is mediated by theinteraction of two plasmid encoded RNAtranscripts, RNA I and RNA II, togetherwith a plasmid encoded protein, Rom orRop, which acts to control plasmid copynumber (1). RNA I and RNA II interactinitially by base-pairing between their com-plementary loop structures (2). Rom bindsto the transiently formed intermediate com-plex, thus suppressing dissociation of thetwo RNAs and facilitating formation of apersistent hybrid duplex of the two RNAs(3), which in turn results in failure of rep-lication initiation. The initial "kissing" (4)

J. D. Marino and R. S. Gregorian Jr. contributed equallyto this work. All the authors are with the Department ofChemistry, Yale University, New Haven, CT 0651 1, USA.*To whom correspondence should be addressed.

interaction between complementary loopshas been proposed as a structural motif ofRNA interaction in other systems (5).

Studies of pairs of hairpins derived fromthe RNA I and RNA II transcripts demon-strated that the hairpins bind to each othersolely through the interaction of their com-plementary loops, and that Rom specificallybinds the structure formed by the interac-tion of these loops (6, 7). In addition, thestability of the complexes formed by theseRNA hairpins varies in ways that do notcorrelate simply with the potential numberand kind of Watson-Crick base pairs thatcould be formed between the complemen-tary loops. The most striking example is thestability enhancement, 350 times greaterrelative to the wild type, observed for com-plexes formed between hairpins in whichthe wild-type loop sequences are inverted

SCIENCE * VOL. 268 * 9 JUNE 19951448

on Decem

ber 8, 2020

http://science.sciencemag.org/

Dow

nloaded from

5' to 3' (6-8). The hairpins with invertedloop sequences dissociate nearly 10,000times more slowly than the wild-type stem

loops. In addition, their association rates

are 10 times slower than those observed forthe wild type (6-8). For both classes it isfound that full base pair complementarityamong the nucleotides of the two constitu-ent loops is required for maximal affinity(6-8).We have characterized by multidimen-

sional heteronuclear magnetic resonance

(NMR) the loop-loop complexes formedbetween RNA hairpins with either comple-mentary wild-type loop or inverted loopsequences (Fig. 1). Full base pairing be-tween the loops was demonstrated in theinverted loop complex by imino spectrosco-py; six of seven potential imino resonances

were observed in the wild-type complex.The base-pair stacking patterns and riboseconformations in both complexes have alsobeen established. While the NMR data donot provide long distance restraints requir-ing a bending of the helical axis of the

RNA II wt

5' 3'

C2 -G18A3 - U17C4- G16C5 -G5G6- C14

C7 A13U8 A12A9 cioC~

U8 A2

U7 G13G6- C14C5- G15C4 G16A U17C2 -G1G. C19

5' 3'

RNA wt

RNA II inv

5' 3'G1..1.C19

C2 -G1A3 U17C4-G16C5 Gi-G6- C14

A7 C13

A U12C9 c 1As

10

AB U12G7 U13

G6 - 4C5 - G15A4 - U16A3 U17C2 -G1G- C19Go ...C20

5' 3'

RNA inv

Fig. 1. The sequence and predicted secondarystructures of the RNA hairpin and RNA II hairpinfor the wild-type (left) and inverted-loop sequenc-es (right). The RNA inverted loop transcript wasdesigned and synthesized with an additional stembase-pair for stability reasons. In addition, the low-er stem sequence was altered to differentiate it

from the wild-type sequence in order to disfavorduplex formation between these hairpins. The

numbering for the RNA inverted loop transcriptbegins with zero to keep the loop numberingschemes consistent between the two sets of mol-ecules (that is, the loop residues are labeled as

positions 7 through 13 in all four stem-loops).

complex, models incorporating the requiredlocal NMR defined geometries were allstrongly bent toward the major groove ofthe helix formed between the loops. Evi-dence for the bent character of the com-

plexes was found by extending the DNAcircular permutation analysis (9) to allowestimation of the extent of bending in thesebimolecular RNA complexes. By analogywith the electrophoretic properties of bentDNA, we conclude that the complex isbent at the locus of the loop-loop interac-tion, and that the bend is further accentu-ated by binding of the Rom protein.NMR of complementary loop-loop

RNA complexes. We synthesized two pairsof RNA hairpin sequences (Fig. 1), witheither the complementary wild-type loopsequences (RNA I wt and RNA II wt) or

complementary inverted loop sequences

(RNA I inv and RNA II inv). Uniformly'5N-labeled RNA hairpins were also syn-

thesized (10). A combination of homo-nuclear 'H and heteronuclear 'H, 15N, and31p correlated NMR methods were used to

define the number and type of intramolec-ular Watson-Crick base pairs formed, thehelical stacking conformation, the ribosesugar pucker and the phosphorus chemicalshifts found in the loop-loop complexes.The NMR spectra of the individual wild-type hairpins have been assigned and thestructural characterization of the hairpinshas confirmed the predicted secondary fold,generally a 6-bp (base pair) stem helix with7-bp loop (1 1). The RNA I wild-type hair-pin, however, was found to contain a distinc-tive, highly structured loop with additionalbase pairs and complete nucleotide stacking(11). This structure may account for themore rapid formation of the wild-type com-

plex, which in turn may be the basis forevolutionary selection of that sequence.

14 131H (ppm)

12

-148

152 -

z

156 tn

-160

Complex formation between hairpinscould be monitored by observing the chang-es in the imino region of the 'H NMRspectra when two hairpins with comple-mentary loops were mixed in the presence

of Mg2+ counterion. The imino spectrawere assigned by two-dimensional (2D) 'H-'5N HMQC (heteronuclear multiple quan-

tum correlation) and by 2D 'H NOESY(nuclear Overhauser exchange spectrosco-py) in H20 (12). To facilitate unambiguousassignment of the imino resonances andtheir NOE (nuclear Overhauser effect) cor-

relates to a given hairpin within the com-

plex, we mixed '5N-labeled hairpins withtheir unlabeled complements. The reso-

nances belonging to the '5N-labeled hair-pin could be selectively observed when i5N-selected or 15N-filtered NMR techniqueswere used.

Eleven resolved imino resonances were

seen in an 'H-' 5N HMQC spectrum (12) ofthe complex formed by the wild-type loopsequences in which RNA I was '5N-labeled(Fig. 2A). Six of these imino resonances

were assigned to the RNA I stem helix andfive resulted from base-pairing in the loop-loop helix, in which the base-pair iminoproton was donated by the RNA I strand.The same region of a 'H-' 5N HMQC spec-

trum taken with the complex formed by theinverted loop sequences, with an 15N-la-beled RNA I hairpin, showed 12 resolvableresonances (imino resonance assignments forG15 and G11 required additional correla-tions found in the 2D NOESY spectrum inH20) (Fig. 2B). Six of these imino resonanc-

es were assigned to the RNA I stem helixand six result from base-pairing in the loop-loop helix, in which the imino proton was

donated by the RNA I strand. The 2DHMQC experiment not only served to dis-perse the 'H spectrum, but also allowed

14 131H (ppm)

12

-148

152 -

a.

z

156 an

160

Fig. 2. An expansion of the imino 1H,15N resonance region of a 2D gradient enhanced 'H-15N HMQCexperiment (12) showing imino resonance correlation for the uniformly 15N-labeled RNA stem-loop

complexed with an unlabeled RNA II stem-loop for the wild-type sequences (A) and the inverted loop

sequences (B). The imino 1H,15N cross-peaks observed in the HMQC are assigned for each complex

spectrum, with the resonances that result from the formation of the base pairs of the loop-loop helix

indicated by arrows.

A sG60 9\

1 G6o % G15

o G16 G18G13 e

U17 UI19 X

IU81 ~~~~~~~~~~~~~~~~~~~~~~~~~~k 1~~~~~~~~~~~~~~~~~~~~~~~~

BG18

GGlOG15 G1 %G18~ G 6

GI G7\X

*U16 U9 U17U12 a e

P

1449SCIENCE - VOL. 268 0 9 JUNE 1995

on Decem

ber 8, 2020

http://science.sciencemag.org/

Dow

nloaded from

potential imino resonances from the loop-loop helix were observed in the complexesformed between hairpins having invertedloop sequences. In the complex formed bythe wild-type loop sequences, one of theseven potential imino proton resonances was

not observed, namely that associated withthe potential base pair between R1U7 and

C

E

v-

1H (ppm)

B

| R liA7

5' 3'

RliU12

l7.0 Fig. 3. (A) The expanded imino-imino 1H NOEcorrelated region of a 2D NOESY spectrum in 90

,RKiAl _ e1_ percent H20, 10 percent D20 (12) of the complexE formed between the RNA and RNA hairpins

1RliCI4 A rll~j3 @ { with inverted loop sequences. Sequential NOE

,~connectivities between the imino resonances ob-

served in the stem helices are indicated by long7

vSand short dashed lines for the R1 and R2 stems,

R espectively; and those NOE connectivities be-@o tween the loop-loop helix base pairs by solid lines.

l ,{s The sequential NOE cross-peaks observed for the

RliGl5 R~iA8 loop-loop helix and stem loop junctions are la-eR21C5 beled. (B) The expanded ribose Hi', pyrimidine

@ @R2iG6 H5 to aromatic (H2,H6,H8) 1H NOE correlatedregion of a 2D NOESY spectrum in 99.996 per-

H (ppm) cent D20 (12) of the complex formed between theRNA and RNA II hairpins with inverted loop se-

quences. Sequential NOEs between H8-H6 (n + 1) and sugar Hi' (n) protons are indicated by a solid linefor the RNA and the RNA II stem-loop strands from the top two base pairs in the stems through the loopnucleotides for the pseudo-continuous helical strand from residues R2iC5 through R1 iG1 5. The criticalsequential NOEs at the stem-loop junction are those between nucleotides R1iC14 and R1iU13 (i,inverted), and also nucleotides R1 iG7 and R2iG6. These connectivities are consistent with a 3' stack ofthe loop nucleotides on their respective stems. NOE correlations between H2 protons and Hi' protonsare labeled in bold, and dashed arrows show that the Hi' chemical shifts are in line with the correctpositions of the traced backbone NOE "walk." (C) Schematic structure of the RNA l-RNA 11 inverted loopcomplex that summarize the NMR results. For clarity only the first two nucleotides in each stem are

drawn. The residues in the remaining portion of the stems were found to display NOEs that were indicativeof a standard A-form geometry. The connection between the sixth and seventh residues in the hairpinsis drawn only to show strand connectivities and has no structural relevance. Base-pair hydrogen bondingis shown by wide solid lines between base-pair rectangles. Ribose Hi', base H8,H6, adenine H2, andimino protons are represented by dots within pentagons, on the outside of the rectangular bases, on theinside of adenines base rectangles, or within base-base hydrogen bonds, respectively. Observed intra-and inter-nucleotide NOEs are indicated by solid lines connecting the dots. Base-pair stacking isindicated by thin, shaded boxes between the stacked base pairs.

R2A13 at one end of the loop-loop helix.Although this imino resonance was not de-tectable, NOE evidence for the stacking ofthe base pair was found in a NOESY exper-

iment in D20 (11, 12).The analysis of the 'H NOESY spectrum

in H20 at 150C (12) indicated that theoverall topology of the wild-type and in-verted loop complexes were quite similar interms of base pair formation and stacking.Since the line widths of the wild-type com-

plex were observed to be broader than themore stable inverted loop complex, indica-tive of intermediate conformational ex-

change, the discussion below is focused on

the inverted loop complex.The imino-imino region of a 2D 'H

NOESY spectrum in H20 (Fig. 3A) of theinverted loop-loop complex showed se-

quential NOE connectivities of the twostem helices and the loop-loop helix. Theseconnectivities established stacking of all ofthe base pairs in the loop-loop helix be-tween the two stem helices and indicatedthat the loop residues had a continuoushelical and stacked geometry on the 3' sideof the respective stem helices. These exper-

iments ruled out the alternative possibilityof continuous helical stacking of the loopon the 9' side of its stem helix. The criticalNOEs that connect the RNA I stem G6imino resonance with the RNA I loop U13imino resonance and that connect theRNA II stem G6 imino resonance with theRNA I loop G7 imino resonance are boxed.

In the complex formed with comple-mentary inverted loops, the stacking of thebase pairs could be further shown by se-

quential NOE correlations in D20 at 250Cbetween the ribose HI' protons and thearomatic H6,H8 and H2 protons (Fig. 3B).The sequential anomeric aromatic NOE"walk" in the expanded plot (Fig. 3B) of theHl',H5 to aromatic H2,H6,H8 region ofthe D2O NOESY of the complex formedbetween inverted sequence hairpins showedthat the loop-loop helix is stacked betweenthe two stem helices. The loop nucleotidesof each stem loop were found to stack in a

continuous helix on the 3' side of theirrespective stem helices, in agreement withthe polarity and continuity of the loop-loophelix deduced from the imino region of theH20 NOESY. In addition to the sequentialanomeric-aromatic NOE walk that con-

nected the helical regions, the correlationof H2 protons with HI' were also indicativeof the stacked helical conformation. In theinverted loop complexes, the A-U base pairsat the first, second, and fifth positions(counting from the RNA I stem) of theloop-loop helix displayed the characteristi-cally strong imino to H2 NOE correlationand also the H2 to Hi' (intrastrand) andHI' (cross strand) NOEs found in A-formhelical geometry. These NOE correlations

SCIENCE * VOL. 268 * 9 JUNE 1995

identification of imino protons by base type(uracil or guanine) because of their charac-teristic '5N chemical shift differences. As-signments made in this way, with further '5Nselective experiments in H20 on complexesformed with the wild-type loop sequenceswhere the RNA II hairpin was uniformly'5N-labeled, indicated that all seven of the

1450

on Decem

ber 8, 2020

http://science.sciencemag.org/

Dow

nloaded from

are indicated in Fig. 3B for the pseudo- calculated models of the inverted loop-loop rated all the NMR connectivity constraints,continuous helical strand formed by resi- complex (Fig. 4A) revealed that the helical which are shown schematically in Fig. 4Cdues R2iC5 through RliG15. A similar axis of the modeled complex was indeed for the inverted loop hairpin complex. InNOE walk can be made for the pseudo- distorted to fit the combined data and to- repeated calculations, all the models thatcontinuous helical strand formed by resi- pological constraints. The model incorpo- were generated had the essential features ofdues R2iG15 through RIC5. A criticalcross strand NOE observed was that be-tween the R2A7 H2 proton and theRIC14 HI' proton. This correlation pro-vided additional evidence for the stem-loop helical stacking at that junction.Taken together, the imino NOE correla-tions and the anomeric-aromatic NOEcorrelations provided decisive evidencefor the formation, coaxial stacking, andpolarity of the loop-loop helix betweenthe two hairpin stem helices. Moreover,the sequential and cross strand NOEs ob-served for the loop-loop helical resonancesindicated that there are only minor distor-tions from helical geometry in the loop-loop interaction.

Analysis of the changes in chemical shiftof the hairpin resonances on binding pro-vided additional evidence that the complexinteraction was a result of the stacking andpairing of nucleotides in the loops. Largechemical shift changes were observed formost loop residue resonances, while onlysmall changes were observed for the reso-nances of the stem helices below the secondbase pair from the apical stem-loop junc-tion. This is consistent with mutational andthermodynamic data indicating that onlythe first two nucleotides in from the loop-stem junction play a role in complex stabil-ity (6-8). Although the backbone geome-try is not well defined for this complex, theH1',H2' region of the COSY (correlationspectroscopy) spectrum (11, 12) indicatesthat all ribose sugars in the complex ap-peared to be predominately in a C3' endoconformation (with the exception of theend residues in the stems). Furthermore,none of the phosphorus resonances are con-spicuous outliers (11, 12) from what wouldbe expected from the standard A-formchemical shift region (13).A three-dimensional model of the com-

plex formed with inverted complementaryloops was calculated with the distance andtorsional constraints derived from theNOESY and COSY experiments (14, 15).Using the protocols in X-PLOR (14), webuilt starting structures with a distance ge-ometry algorithm and then refined them bya series of restrained molecular dynamicscalculations (15). Although the NMR datamainly define local geometries and do notprovide direct evidence for a global bendingof the overall helical axis of the complex,the requirement for coaxial stacking of allseven base pairs between the stem heliceswas quite suggestive of a bending distortionthat must occur to accomplish this overallfold. A stereoview of one of the refined

Fig. 4. (A) Stereo view of the three-dimensional NMR-derived model constructed for the invertedloop-loop complex. The model was generated with the distance geometry and restrained moleculardynamics protocols in X-PLOR (14, 15). The RNA inverted hairpin is in yellow and the RNA II invertedhairpin is in blue. The model shows the base-pairing and -stacking observed in the complex and thepronounced bend that is a result of the required stacking and helical geometries. The R2iG6, R2iA7 andthe R1 iG6, R1 iG7 dinucleotide steps are shown in green and red, respectively. (B) Stereo view of thethree-dimensional NMR-derived model with the continuous helical stacking observed in the complexshown by color coding the residues that make up the two pseudo-continuous strands of the distortedhelix formed by the RNA hairpins. Residues R2iG1 through R2iG6 and R1 iG7 through R1 C19 are shownin yellow; residues R2iC19 through R2iA7 and R1 iG6 through R1 iG1 are shown in blue. (C) Stereo viewof the three-dimensional model showing the compressed groove and the "bridging" phosphodiesterbackbones (with "bridging" phosphate atoms shown in red) that connect residues R1G6, R1G7 andR2G6, R2A7. The RNA inverted hairpin is in yellow and the RNA II inv hairpin is in blue. Ribbons are

drawn to show the trace of the phosphodiester backbone of the hairpins.

81 140

1451SCIENCE * VOL. 268 - 9 JUNE 1995

on Decem

ber 8, 2020

http://science.sciencemag.org/

Dow

nloaded from

a sharply bent structure with continuousstacking of the base pairs. However, becauseof the limited number of constraints, thestructural model is under-determined andtherefore a significant variation in localgeometry was observed between structures.

The calculated model suggested thatcoaxial stacking of the 7-bp loop-loop he-lix was accomplished under the constraintof closing the loop by bending stronglytoward the major groove. This bend short-ened the distance required for bridgingacross the major groove. The loop-loophelix formed a rather regular helical basestacking geometry, providing for a pseudo-continuous helical strand from one stemhelix through the loop-loop helix to theother stem helix (Fig. 4B).

In the model, the backbone phosphodi-ester linkages between the sixth and sev-enth nucleotides on the 5' side of the hair-pin loops (Fig. 4, A and C) were adjacent insequence, but 7 bp away in the helicalstructure. Their linkage is sterically feasiblebecause of the helical periodicity that plac-es these phosphates directly adjacent acrossthe narrowed major groove of the A-formhelix (Fig. 4C). This geometry is a directconsequence of the 3' stacking of the loopresidues on their respective stem helices. Ifthe loop residues were 5' stacked on theirstems, the 3' residues which would have tobe linked to close the loop would need tobridge through a much larger distance.

Gel electrophoretic analysis of theloop-loop complex. Bending, a well-estab-lished DNA conformational response to se-quence change or protein-ligand binding,can be characterized by gel electrophoresismethods (16), which have also been used todetect bending of RNA (17, 18). To deter-mine the existence and extent of helicaldistortion present in the loop-loop com-plex, and the extent to which the bindingof the Rom protein to the RNA complexeffects further helical distortion, we used amodification of the DNA circular permuta-tion assay (9) to measure RNA-RNA com-plex bending. Complexes formed betweenstem loops with the loop-loop interaction atthe center of a pseudo-continuous helixshowed anomalously slow mobility in anondenaturing polyacrylamide gel whencompared with RNA complexes of similarlength with the loop-loop interaction nearthe end of the molecule; this is the expectedresult if the loop-loop interaction creates abend in the continuous helix.

The circular permutation RNA hairpinprobes, containing either the RNA I orRNA II inverted complementary loop se-quences (Fig. 5A), of systematically variedlengths were made by run-off transcriptionwith T7 RNA polymerase from DNA tem-plates (119). In all complexes, the RNAhairpin probes were mixed so that the total

1452

number of base pairs of RNA that can beformed is fixed at 126 for the total of thetwo stems, plus 7 for the loop-loop helix.Both the RNA complex and the RNA com-plex bound by the Rom protein (20) ana-lyzed by this method (Fig. 5B) displayedposition-dependent variations in electro-phoretic mobility (21). Moreover, the posi-tion-dependent change in relative mobili-ties observed in the two gels indicated thatthe bend found for the RNA loop-loopcomplex was further enhanced by Rombinding.

ARNA probes:

Rlicyc R2icyc IG10/IC10(1 ) s, L-,

T7bsfT7block

AI L1 L n L

(2) L I_

(3)l -n _ _

(4)l L o _M_n LnL-,9

(5) n L n9_

(6)i 9 C 9 ; 9 _ I

(7) | -n n I 9 9 L"

(8) | -

(9) 1 9 )I n L I L L"

(10)l Lo E

< 130 bp

Although the orientation of the bendcould not be ascertained by this circularpermutation assay, the extent of bendingcould be roughly estimated by a comparisonof the gel mobilities of the circularly per-muted complexes with the gel mobilities ofsimilarly permuted standardized constructsof known curvature, for which we usedDNA fragments of 120 bp containing 2 or 3A-tract sequences (22). (A slightly shorterDNA sequence length was chosen in partialcompensation for the smaller rise per basepair for helices of the A-form as compared

BRNA alone

1 2 3 4 5 6 7 8 9 10 S

A-tract constructs:

. 120 bp 0

Fig. 5. Circular permutation analysis (9) of c

the distortion in the inverted stem-loop 1.05complex. (A) Schematics of the RNA andA-tract probes used in the circular permu- 1tation assay (8, 19). The RNAs contain the

10-bp linker sequence IG101C10, the

T7bs-T7block duplex (which contains the 0 0.9 *

T7 RNA polymerase promoter in the cor- E ....*0

responding DNA sequence), and either the k 0.85

R1icyc or R2icyc hairpin sequences. The 2tracts.2~~~~~~~~£1 0. RA2 r6t

probes used to generate the two- and 0.8 RNA a

three-phased A-tract molecules for bend 075 A6tracts

calibration were generated by restriction RNA+ Rom

endonuclease cleavage at sites designed 0.71 6 80 1 120at 1 0-base intervals within the two flanking

DNA seuences(22).(B) ElctrophreticBase pairs from end of circular permuted probeDNA sequences (22). (B) Electrophoretic *mobility-shift analysis of the complexes [1 through 10 from the schematic in (A)] formed between the RNAand RNA 11 hairpins with inverted-loop sequences bound to circularly permutated probes. The lane

marked with an "S" was a single RNA probe containing 128 bp plus the RI loop. Complex formationbetween a given pair of hairpins and mobility modulation by cyclic permutation of the loop-loop interactionsite was shown with nondenaturing polyacrylamide gel electrophoresis. The same assay was performedafter incubating the Rom protein with the loop-loop complexes formed between the same RNA and RNA11 hairpins bound to circularly permutated probes. (C) Calibration curve for estimation of the extent ofbending of the loop-loop complex. The mobilities of the RNA complex, the RNA plus Rom complex, andthe A-tract standards have been fit to cosine curves and are plotted as a function of position of the bendlocus from the end of one of the circular permuted probes (23).

SCIENCE * VOL. 268 * 9 JUNE 1995

RNA + Rom

. -M-11---. om

bo

MMA- -W_~- qFwWROto

on Decem

ber 8, 2020

http://science.sciencemag.org/

Dow

nloaded from

SSgAXr SMMc"

to those of the B-form.) From analysis ofthese results (23), we estimated a bend an-gle for the loop-loop complex of 450 and abend angle for the loop-loop-Rom complexof 600. This calibration was only approxi-mate because it relied on comparison ofmolecules of different shape and potentiallydiffering flexibility. The position of the mo-bility minimum indicated that the bends ofthe loop-loop RNA complex and the RNA-Rom complex were centered at the loop-loop interaction locus.

Elements of RNA-RNA recognition.Base-pairing between complementary re-gions and coaxial stacking of helices arecommonly observed tertiary interactions inRNA structures (24). The loop-loop com-plex described above folds the complemen-tary loop sequences of two independentRNA secondary elements in a pseudoknot-like (25) fashion. Complex formation is driv-en by the base-pairing between complemen-tary loops and stacking of the loop-loop helixbetween the stem helices of the respectivehairpins. As has been observed in pseudo-knotted RNA structures (26), the stackingenergy contributes to the binding energy ofthe helical interaction. In these hairpin loopcomplexes, differential stacking energies atthe helix-helix junctions may contribute tothe stability difference between wild typeand inverted loop complexes.

The overall bend of the complex can beviewed as a compensatory structural changenecessary for complete base-pairing and co-axial helical stacking. The energetic cost ofthe required backbone helical distortions ismore than compensated for by the energeticgains from base-pair formation and stack-ing. Stable loop-loop complexes are usuallyformed between hairpins with loop sizes ofsix to eight nucleotides. As Pleij and co-workers pointed out (25), the RNA A-formmajor groove width to be spanned by one ofthe pseudoknot loops is at a minimum (10.1A) when the pseudoknot helix contains sixto eight base pairs. This contrasts with thewidth of the minor groove that must bespanned by the second pseudoknot loop,which increases approximately linearly withincreasing size of the pseudoknot helix.Complexes formed between hairpins withlarger loop sizes or with 5'-stacking of loopresidues on their respective stems, couldonly achieve a similar coaxially stackedstructure at a much greater energetic cost todistort the RNA backbone.

Even though the complexes fold in orderto minimize the distances required for loopclosing across the major groove, the bendthat results from complex formation (-45°to 90°) is still quite large for such a shortRNA sequence. However, the large bend inDNA observed in the TBP complex givesprecedent for such a structure (27); bendingin that case also results from narrowing the

major groove and widening the minorgroove. The sharp bend in the model struc-ture also bears an overall resemblance totRNA, an observation that may be of inter-est in view of the conjecture that tRNAmight have evolved in two stages or fromtwo separate smaller RNAs (28). Combininga hairpin having a -CCA 3' amino acidacceptor with a dumbbell shaped moleculeincorporating an anticodon loop would pro-vide an appropriate overall tRNA-like ge-ometry in their loop-loop complex. In theColEl system, the bending of the loop-loopcomplex may play a functional role in bring-ing together distal regions of the larger RNAI and RNA II transcripts and thereby facili-tate further folding after the initial loop-loopinteractions.

Implications for Rom recognition. Romis a four helix-bundle protein whose struc-ture has been characterized both by diffrac-tion (29) and NMR methods (30). Muta-tional analysis of the solvent exposed aminoacids has revealed that the important resi-dues for RNA binding are all located onhelix I of the helix-loop-helix monomersand are consequently all found on a singleface of the antiparallel four-helix dimer(31). Nonetheless, the key to Rom's speci-ficity for the loop-loop structure over gener-ic A-form RNA is not apparent in theprotein structure alone. The key recogni-tion element, however, could reside in thebent helical RNA structure created by theinteraction of the two hairpins rather thanin recognition of a specific nucleotide se-quence. The ability of complementary loopsto form the distorted helical fold seems tobe the primary requirement for specific rec-ognition. In addition, the phosphodiesterresidues that bridge the compressed majorgroove of this RNA complex provide apotentially attractive sequence-neutral rec-ognition element for Rom binding.

The prior organization of this RNAcomplex into a form that favors proteinbinding contrasts with the Tat-TAR inter-action, in which the RNA structure is re-organized upon binding peptides taken fromthe HIV transactivating or Tat protein tothe HIV RNA sequence target TAR (32).It can be inferred that the RNA bendingdistortion is an important element in theprotein-RNA complex structure, since Rombinding increases the bend angle beyondthat seen in the RNA loop-loop complex.This observation provides a possible expla-nation for the preference of Rom for theloop-loop complex over normal A-formRNA, since prior bending of the RNA inthe loop-loop complex should reduce thefree energy required for the distortion rela-tive to the A-form helix. This is analogousto enhanced binding affinity observed forDNA bending proteins when the DNAbinding site has already been bent (33).SCIENCE * VOL. 268 * 9 JUNE 1995

REFERENCES AND NOTES

1. Y. Eguchi, T. Itoh, J. Tomizawa, Annu. Rev. Bio~chem. 60, 631 (1991); R. W. Simons and N. Kleck-ner, Annu. Rev. Genet. 22, 567 (1988).

2. J. Tomizawa, J. Mol. Biol. 212, 683 (1990).3. ibid., p. 695.4. ,Cell 38, 861 (1984).5. K.-Y. Chang and 1. Tinoco Jr., Proc. Natl. Acad. Sci.

U.S.A. 91, 8705 (1994); C. Persson, E. G. Wagner,K. Nordstrom, EMBOJ. 9,3767 (1990); H. Leffers, J.Kjems, L. Ostergaard, N. Larsen, R. J. Garrett, J.Mol. Biol. 195, 43 (1987); P. J. Cascone, T. F. Hay-dar, A. E. Simon, Science 260, 801 (1993).

6. Y. Eguchi and J. Tomizawa, J. Mol. Biol. 220, 831(1991).

7. R. S. Gregorian Jr. and D. M. Crothers, J. Mol. Biol.,in press.

8. R. S. Gregorian Jr., thesis, Yale University (1994).9. H. M. Wu and D. M. Crothers, Nature 308, 509

(1984).10. All NMR samples were synthesized by T7 run-off

transcription (34) and purified by standard gel elec-trophoresis methods. Unlabeled samples were syn-thesized with commercially available NTP (nucleotidetriphosphates). The 15N-labeled NTPs were pre-pared from RNA isolated from E. coli grown in mini-mal media with 15NH4C1 as the sole nitrogen source(35). NMR experiments were performed with RNAsample concentrations of 1.0 to 1.5 mM (with theexception of the 15N-labeled RNA inverted loopsample, which was 0.4 mM) in 5 mM Mg2Cl, 50 mMNaCI, and 1 mM cacodylate, pH 6.5.

11. J. P. Marino, thesis, Yale University (1995).12. The exchangeable imino, amino, and aromatic H2

and H5 protons for the hairpins and complexes were

assigned by means of 300-ms gradient-enhancedNOESY experiments in 90 percent H20 with 10 per-cent D20. The spectra were collected at 1 50C withthe use of a standard NOESY pulse sequence inwhich the last 900 pulse was replaced by a gradient-enhanced 1 -1 spin echo pulse sequence (36). The2D NOESY experiments spectra were acquired with256 and 1024 complex points in t 1 and t 2 with 64scans per t 1,t 2 increment and spectral widths of10,000 Hz in each dimension. Gradient-enhanced1H-15N HMQC (36) and 15N-filtered 300-ms NOESYexperiments were also collected in 90 percent H20,10 percent D2O at 1 50C with selective off-resonanceexcitation accomplished with gradient-enhanced1-1 spin echo pulses incorporated in the context ofthe proton pulses of the experiment (36). The HMQCspectra were acquired with 64 and 512 complexpoints in t, and t2 with 16 scans per t1,t2 incrementand spectral widths of 1600 and 10,000 Hz, respec-tively. The 15N-filtered NOESY experiments [G. Ot-ting and K. Wuthrich, 0. Rev. Biophys. 23, 39 (1990)]were acquired with 128 and 512 complex points in t1and t2 with 64 scans per t1 ,t2 increment and spectralwidths of 10,000 Hz in each dimension. These ex-

periments were collected on a GE Omega 500 spec-trometer with x,y, z gradients applied for 1 ms duringthe spin echo delays in the NOESYs and during de-focusing-refocusing periods of the HMQCs to elimi-nate transverse H20 magnetization not refocused bythe selective 180° pulse. Assignment of the nonex-

changeable aromatic H2, H5, H6, H8, and the Hi',H2' proton resonances was made with NOESY,COSY, DQF-COSY, and 1 H 1 5N HMBC experimentscollected in 99.996 percent D20 performed on eithera GE Omega 500 or a Bruker AM500 spectrometerat 25°C. The 2D NOESY and COSY experimentswere collected with standard pulse sequences [K.Wuthrich, NMR of Proteins and Nucleic Acids (Wiley,New York, 1986)]. For both experiments, 256 and1024 complex points in t1 and t2 were acquired with32 scans per t1,t2 increment. The NOE mixing time(tm) was 300 ms. The 'H-15N refocused HMBC ex-

periment was collected with 128 and 512 complexpoints in t1 and t2 with 64 scans per t t2 increment.Sequential NOEs from anomeric ribose Hi' protonsto aromatic H6-H8 protons were established withthe NOESY; pyrimidine base H5,H6 and ribose Hi',H2' correlated cross-peaks were identified by theCOSY, and aromatic resonances (H2,H5,H8) were

identified by correlation to specific base nitrogens in

1453

on Decem

ber 8, 2020

http://science.sciencemag.org/

Dow

nloaded from

the refocused HMBC. One-dimensional phosphorusspectra were taken to determine the chemical shiftranges for these resonances in the complexes; how-ever, phosphorus correlated experiments were notused for assignment of the complexes. All spectrawere processed on an Indigo workstation (SiliconGraphics) with Felix software (Biosym Technologies,San Diego, CA).

13. P. Legault and A. Pardi, J. Magn. Reson. Ser. B 103,82 (1994).

14. A. Brunger, X-PLOR, A System for X-ray Crystallog-raphy and NMR (Yale Univ. Press, New Haven, CT,1992).

15. Models for the loop-loop complex were calculated bymeans of distance geometry and molecular dynam-ics routines in X-PLOR (14) and displayed with Insight11 (Biosym Technologies). Restraints used in the X-PLOR calculation included exchangeable NOEs fromimino to imino and from imino to amino and H2 (ad-enine) resonances that were taken from the H20NOESY experiments (observed exchangeable protonNOEs were characterized as 3.5 ± 1.0 A) and NOEsfrom the Hi' to aromatic H6, H8, and H2 resonancestaken from the D20 NOESY experiment [nonex-changeable proton NOEs were characterized asstrong (2.4 + 0.5 A), medium (3.0 ± 1.0 A), and weak(4.5 + 1.5 A)]. Some H2' to aromatic H6-H8 NOEswere identified, but this region of the spectrum dis-played much greater chemical shift overlap and sowas interpreted in only a limited way. In the lowerstem regions, the first four base pairs of both the RNAand RNA II hairpins, the bond angles, and distance

restraints used in the calculations were taken from W.Saenger [Principles ofNucleicAcid Structure (Spring-er-Verlag, New York, 1988)] for a standard A-formhelix. Experimental NOE restraints in this region in-cluded 91 inter-residue and 36 intra-residue con-straints. In the upper two base pairs of each helix andin the loops there were a total of 64 inter-residue and44 intra-residue constraints. Pseudo-NOEs wereused to define the hydrogen bonding network for allbase pairs in the complex, with a total of six proton-heavy atom constraints for each GC pair and four foreach AU pair. To define the ribose sugar pucker, theabsence of ribose Hi -,H2' cross-peaks in the COSYexperiment for all but the terminal bases in the stemswas used to constrain the sugar puckers as 03'-endo (in a C3-endo conformation, 3J(H1 ,H21) <2 Hz). All bases in the loop-loop complex were re-quired to be in the anti conformation and a smallenergy term (50 kcaVA2) was included to limit thefreedom of base pairs to depart from planarity. Allbackbone angles were loosely constrained (- 1 5° inthe lower four base-pair residues of the hairpin stemsand +300 forthe top two stem base-pair and the loopresidues) to lie in the expected A-form geometry in thefirst round of distance geometry embedding and sim-ulated annealing, after which those structures withthe correct loop-loop topology (that is, structureswithout catenated loops) were further refined with thebackbone angular restraints completely relaxed forthe top two stem base-pair and "loop" residues. Theloop-loop structures were refined by several roundsof simulated annealing by first fitting them to a tem-plate coordinate set to correct local geometry. Eachstructure was then heated to 2000°K and the NMRrestraints were applied. The bond, angle, and im-proper energy terms were simultaneously included.After equilibrating the molecule for 6 ps at high tem-perature, a repulsive Van der Waals term was turnedon, and each structure was slowly cooled to 1 00K.No electrostatic term was included in the calculation.

16. Reviewed by D. M. Crothers, M. R. Gartenberg, T. E.Shrader, Methods Enzymol. 208,118 (1991); D. M.Crothers and J. Drak, ibid. 21 2B, 46 (1992).

17. A. Bhattacharyya, A. I. Murchie, D. M. Lilley, Nature

343, 484 (1990).18. R. S. Tang and D. E. Draper, Biochemistry 29, 5232

(1 990).19. The DNA templates were built from the following con-

stituent pieces (8): (i) T7b, which contains the T7promoter bottom strand; (ii) T7block, which annealsto T7b and facilitates its attachment to multimers; (iii)IG-10, IC-10, and their tandem repeats IG-20 andIC-20, which are linkers of 10 or 20 nucleotides, re-spectively, in length with 2-nt overhangs; and (iv)Riicyc and R2icyc templates encoding the hairpinloop sequences. DNAs were synthesized by stan-dard phosphoramidite chemistry on an Applied Bio-systems 380B DNA synthesizer, and phosphorylatedat their 5' ends with T4 DNA polynucleotide kinase(New England Biolabs) in the commercial buffer, sup-plemented with additional ATP to a final concentra-tion of 10 mM, for 60 minutes at 37°C. T7b was firstlabeled at the 5' end by incubation with [y-32P]ATP(New England Nuclear, 6000 Ci mmol- 1) andT4 DNApolynucleotide kinase for 30 minutes followed by anadditional incubation with unlabeled ATP (added to afinal concentration of 10 mM) for 30 minutes. IG-10and IC-10, IG-20 and IC-20, and T7b and T7blockwere then mixed at equimolar concentration, heatedto 900C and slowly cooled to hybridize the comple-mentary DNA strands. DNA template componentswere then ligated in two stages with T4 DNA ligase inthe commercial buffer supplemented with ATP to a2.5 mM final concentration. In the first ligation, theT7b-T7block hybrid was mixed with either the IG-IC-10 or IG-IC-20 hybrids, or with a mixture of both at a1 :10 molar ratio. The final mixture was ligated (over-night at 1 5°C), precipitated with ethanol, and fraction-ated on an 8 percent nondenaturing polyacrylamidegel in a buffer of 89 mM tris, 89 mM boric acid, pH8.3, and 0.1 mM EDTA. The observed ligation ladderconsists of the T7b-T7block duplex ligated onto in-creasing numbers of the IG-IC-1 0 or -20 monomers.These multimers were excised from the gel, eluted at370C overnight, and precipitated with ethanol. Eachof the excised DNA multimers was then "capped"with one of the "cyc" end-pieces to make a hairpintemplate. The cyclic endpiece was added in excess(about four times more) to linear duplex DNA, and thepieces were ligated as above. Reactions werequenched at 65°C. The ethanol-precipitated prod-ucts were used for transcription by T7 RNA poly-merase without further purification [a-32P]UTP (10,uCi) was added to transcription mixtures to producebody-labeled RNA hairpins. Reactions were gener-ally incubated overnight (8 to 12 hours) at 37°C andpurified by standard methods on either 6 percent(longer RNAs) or 12 percent (shorter RNAs) denatur-ing polyacrylamide gels.

20. Rom protein was purified as described [R. M. Lacat-ena, D. W. Banner, G. Cesareni, in Mechanisms ofDNA Replication and Recombination, N. Cozarelli,Ed. (Liss, New York, 1983), pp. 327-336] as opti-mized in our laboratory from E. coli cells obtainedfrom B. Polisky [M. Muesing, C. D. Carpenter, W. H.Klein, B. Polisky, Gene 31, 155 (1984)].

21. For each complex, separate RNAs were individuallyplaced in 5 mM MgCI2, heated to 90°C, and cooledimmediately on ice. The R1 and R2 for each reactionwere then mixed on ice, incubated at 15°C for 20minutes, and then transferred to 15 percent poly-acrylamide gels (75:1, acrylamide:bis-acrylamide,0.75 mm thick) and separated in a Hoeffer constanttemperature gel apparatus, at 300 V (<5 W per gel)until the xylene cyanol was at the bottom third of thegel (14 hours). The running buffer was 89 mM tris, 89mM boric acid, pH 8.3, and 5 mM MgCI2. The bandswere visualized by autoradiography.

22. The reference DNA molecules were constructedfrom three separate DNA duplexes and contained

either an A6-N4-A6 or an A6-N5-A6-N4-A6 sequence(9). Molecules with two- or three-phased A6 tractswere constructed from their three constituent partsby exhaustive ligation of the components that hadbeen synthesized, purified, and phosphorylated asabove (19). Reference DNA molecules were subject-ed to electrophoresis under identical conditions tothe loop-loop complexes (21). All reference mole-cules were heated to 55°C, cooled on ice, and trans-ferred to the gels. The two- and three-phased A6tracts bend the DNA by 360 and 540, respectively,which nearly brackets the range of mobility modula-tion observed from the loop-loop complexes.

23. Gel autoradiographs were analyzed for mobility ofthe individual bands for each complex or A-tractstandard by measuring the distance from the bottomof the wells to the middle of the band. The data werefit to a cosine curve to evaluate the projected mobilityof the complex when the bend center was at the endof the DNA molecule. From this number (pL), the rel-ative mobility (urei) of each bend position was calcu-lated. Comparison of the relative mobilities of theloop-loop complexes with those of the A-tract bendsof known magnitude allowed estimation of the bendangle of the complex, if we assume that there is aquadratic dependence of the position-dependentanomaly on curvature (16).

24. M. Chastain and 1. Tinoco Jr., Prog. Nucleic AcidRes. Mol. Biol. 41, 131 (1991); J. R. Wyatt and 1. J.Tinoco, in The RNA World (Cold Spring Harbor Lab-oratory Press, Plainview, NY, 1993), pp. 465-497.

25. C. W. A. Pleij, K. Rietveld, L. Bosch, Nucleic AcidsRes. 13, 1717 (1985).

26. J. R. Wyatt, J. D. Puglisi, I. Tinoco Jr., J. Mol. Biol.214, 455 (1990); J. D. Puglisi, J. R. Wyatt, I. TinocoJr., Acc. Chem. Res. 24, 152 (1991).

27. Y. Kim, J. H. Geiger, S. Hahn, P. Sigler, Nature 356,512 (1993); J. L. Kim, D. B. Nikolov, S. K. Burley,Nature 356, 520 (1993).

28. N. Maizels and A. Weiner, Proc. Natl. Acad. Sci.U.S.A. 91, 6729 (1994); P. Schimmel and B. Hen-derson, ibid., p. 11283.

29. D. W. Banner, M. Kokkinidis, D. Tsemoglou, J. Mol.Biol. 196, 657 (1987).

30. W. Eberle, W. Klaus, G. Cesareni, C. Sander, P.Rosch, Biochemistry 29, 7402 (1990); W. Eberle, A.Pastore, C. Sander, P. Rosch, J. Biomol. NMR 1, 71(1991).

31. L. Castagnoli etal., EMBOJ. 8, 621 (1989).32. J. D. Puglisi, R. Tan, B. J. Caman, A. D. Frankel, J. R.

Williamson, Science 257, 76 (1992); J. D. Puglisi, L.Chen, A. D. Frankel, J. R. Williamson, Proc. Natl.Acad. Sci. U.S.A. 90, 3680 (1993).

33. J. D. Kahn and D. M. Crothers, Proc. Natl. Acad. Sci.U.S.A. 89, 6343 (1992).

34. J. F. Milligan, D. R. Groebe, G. W. Witherell, 0. C.Uhlenbeck, NucleicAcids Res. 15,8783 (1987); J. F.Milligan, 0. C. Uhienbeck, Methods Enzymol. 180,51 (1989).

35. R. T. Batey, M. Inada, E. Kujawinski, J. D. Puglisi, J.R. Williamson, Nucleic Acids Res. 20, 4515 (1992);E. P. Nikonowicz et al., ibid., p. 4507.

36. V. Sklenar and A. Bax, J. Magn. Reson. 74, 469(1987); A. A. Szewczak, G. W. Kellogg, P. B. Moore,FEBS Lett. 327, 261 (1993).

37. We thank G. Sun for assistance with oligonucleotidesynthesis; B. Polisky (Indiana University) for providingE coli cells from which Rom protein was purified;and J. H. Prestegard for advice and suggestions (toJ.P.M.) on the NMR experiments. Supported by theNational Institutes of Health (NIH) (GM 21966)(D.M.C.), an NIH Training Grant in Biophysics(J.P.M.), and a Howard Hughes Medical InstitutePredoctoral Fellowship (R.S.G.).

7 November 1994; accepted 3 April 1995

SCIENCE * VOL. 268 * 9 JUNE 19951454

on Decem

ber 8, 2020

http://science.sciencemag.org/

Dow

nloaded from

Bent helix formation between RNA hairpins with complementary loopsJP Marino, RS Gregorian Jr, G Csankovszki and DM Crothers

DOI: 10.1126/science.7539549 (5216), 1448-1454.268Science 

ARTICLE TOOLS http://science.sciencemag.org/content/268/5216/1448

REFERENCES

http://science.sciencemag.org/content/268/5216/1448#BIBLThis article cites 39 articles, 7 of which you can access for free

PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions

Terms of ServiceUse of this article is subject to the

is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title (print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience

No claim to original U.S. Government Works.Copyright © 1995 The Authors, some rights reserved; exclusive licensee American Association for the Advancement of Science.

on Decem

ber 8, 2020

http://science.sciencemag.org/

Dow

nloaded from