Proc. Natl. Acad. Sci. USAVol. 73, No. 7, pp. 2294-2298, July 1976Biochemistry

Hybridization of RNA to double-stranded DNA: Formation ofR-loops

(gene isolation/gene mapping/electron microscopy/restriction endonucleases)

MARJORIE THOMAS, RAYMOND L. WHITE, AND RONALD W. DAVIS

Department of Biochemistry, Stanford University School of Medicine, Stanford, California 94305

Communicated by Norman Davidson, April 26,1976

ABSTRACT RNA can hybridize to double-stranded DNAin the presence of 70% formamide by displacing the identicalDNA strand. The resulting structure, called an R-oop, is fonnedin formamide probably because of the greater thermodynamicstability of the RNA-DNA hybrid when it is near the denatura-tion temperature of duplex DNA. The rate of R-loop formationis maximal at the temperature at which half of the dIlexDNAis irreversibly converted to single-stranded DNA (the strandseparation temperature or ts) of the duplex DNA and fallsprecipitously a few degrees above or below that temperature.This maximal rate is similar to the rate of hybridization of RNAto single-stranded DNA under the same conditions. At temper-atures above the t5s the rate is proportional to the RNA con-centration. However, at temperatures below t55 the rate of R-loop formation is less dependent upon the RNA concentration.Once formed, the R-loops display considerable stability; theformamide can be removed and the DNA can be cleaved withrestriction endonucleases without loss of R-loop structures.

It has recently been observed by R. L. White and D. Hogness(manuscript in preparation) that rRNA can hybridize to duplexDNA that codes for rRNA (rDNA) of Drosophila melanogaster.As observed in the electron microscope, these R-loop structuresappear similar to the D-loop structures reported by Robbersonet al. (1). However, the duplex portion of the loop is an RNA-DNA hybrid rather than a DNA-DNA hybrid and has thus beentermed an R-loop. In order to optimize the formation and uti-lization of R-loops observed by White and Hogness, we havestudied their rate of formation and kinetic stability under avariety of conditions. The ability to form R-loops at high effi-ciency under controlled conditions may facilitate the mappingand isolation of DNA sequences complementary to specificRNAs.

MATERIALS AND METHODSReagents. Xgt-Scl109 DNA and yeast total rRNA were

generously supplied by R. Kramer. Formamide was purifiedby crystallization at -3' in a salt-ice bath. The mixture wasstirred with a motor-driven, stainless steel paddle to the con-sistency of soft ice cream. A 250-ml polycarbonate centrifugetube was cut in half and a nylon screen cemented between thehalves. The formamide crystals were recovered by centrifu-gation onto the nylon screen. A hole, drilled just below thescreen, was used to remove the sedimented liquid. The form-amide was stored at -20°. For hybridization reactions, smallglass test tubes (5 X 60 mm) were siliconized by treatment witha 1% solution of dimethyl dichlorosilane dissolved in benzene.The tubes were then dried in an oven of 200° and extensivelywashed with water.

R-loop Formation Buffer. A buffered formamide solutionwas first prepared by mixing 0.42 ml of formamide, 50 Al of 1M Pipes [piperazine-NN'-bis(2-ethanesulfonic acid) Naj.4] atpH 7.8, 12 Al of 0.5 M Na3EDTA, and 18 Ai of H26. Fifty mi-croliters of this solution were delivered with Teflon or poly-ethylene tubing to the bottom of a 5- X 60-mm siliconized glasstest tube. Ten microliters of a solution containing approximately30 Ag/ml each of RNA and DNA, 0.1 M NaCl, and 0.05 MTris-HCl at pH 7.5, was then added. The final cation concen-tration is approximately 0.17 M. The solution was covered withparaffin oil and the tube sealed with Parafilm. The test tubeswere incubated in a Haake constant temperature bath filledwith ethylene glycol.

Determination of Strand Separation Temperature (t,.).DNA cleaved with EcoRI endonuclease was placed in the R-loop formation buffer as described above. Samples of 4 AI eachwere taken at 10 intervals starting at 45°. Five minutes wereallowed for equilibration between intervals. Each sample wasdiluted 20-fold into 0.35 M ammonium acetate at pH 8, 0.01M Na3EDTA, and 75 Mg/ml of cytochrome c at 00. Sampleswere mounted for electron microscopy by the aqueous dropmethod. Drops of 25 Ml of the above DNA plus cytochromesolutions were placed on a polished teflon bar. A parlodion-coated microscope grid was touched to the side of each dropand was stained with uranyl acetate (2). At the strand separationtemperature (tss) of a specific EcoRI DNA fragment, the duplexstrands are converted into collapsed single-strand bushes. Thetss of the RNA-DNA hybrid was determined in an identicalmanner, except that RNA was first hybridized to its EcoRI-cleaved complementary DNA strand at 470 in the R-loop for-mation buffer.Our initial work on R-loop formation was complicated by

DNA degradation and nonreproducible reaction rates. Webelieve this was the consequence of changes in the reactionconstituents and their concentrations during the course of thereaction. The following revisions have resulted in our abilityto obtain reproducible R-loop formation rates with no detect-able DNA degradation when assayed by electron microscopy:(1) recrystallized formamide was used; (2) the reactions wereconducted under oil in sealed siliconized glass test tubes, whichprevented evaporation and solvent condensation on the walls;(3) the temperature was maintained to within -0.20; and (4)the decomposition of the RNA and the pH change of theformamide solution was minimized by using Pipes buffer at pH7.8.

RESULTSModel system for R-loop formationThe study of R-loop formation has been greatly facilitated bythe use of a simple model system. A viable hybrid DNA mole-cule (Xgt-Sc1109) containing the 2.6 kb (kilobase pair) repetitive

2294

Abbreviations: Pipes, piperazine-N,N'-bis(2-ethanesulfonic acid)-Nal.4; tss, temperature at which half of the duplex DNA is irreversiblyconverted to single-stranded DNA; tm, temperature at which 50% ofthe base pairs are no longer paired; tmax, temperature at which maxi-mum reaction rate occurs; rDNA, DNA that codes for rRNA.

Dow

nloa

ded

by g

uest

on

Aug

ust 8

, 202

1

Proc. Nati. Acad. Sci. USA 73 (1976) 2295

Table 1. Strand separation temperatures (t55)of EcoRI segments of Xgt-ScllO9 DNA

EcoRI Approximatesegment tS8 0C % G+C (5, 6)

XB 45 37rDNA 49 47xgt right 50 47xgt left 55 57rDNA rRNA 58 47

EcoRI DNA fragment from Saccharomyce cerevlsiae (yeast)covalently inserted into the middle of bacteriophage X DNAwas used. The inserted DNA fragment hybridizes to yeast 26SrRNA, which can easily be prepared from yeast cells in largequantities. This X-yeast hybrid is one of several isolated fromlarge pools of A-yeast hybrids by in situ hybridization of RNAto individual plaques (Kramer, Cameron, and Davis, manu-script in preparation). As shown below, the yeastDNA fragmentcoding for rRNA (rDNA) is about T4 of the length of the 26SrRNA and hybridizes at one end of the 26S rRNA.

Strand separation temperatures of the Xgt-Sc EcoRIDNA fragmentsThe strand separation temperatures of the EcoRI DNA frag-ments from Xgt-ScI109 were determined under the R-loopreaction conditions (70% formamide, 0.1 M Pipes, 0.01 M-NasEDTA). Samples were taken at 10 intervals starting at 450and examined in the electron microscope as described in Ma-terials and Methods. The temperature at which half of theduplex DNA is irreversibly converted to single-stranded DNAis defined as the strand separation temperature (t.) of thisfragment. The strand separation of the rDNA fragment oc-curred within a 10 interval, while the strand separation of theleft and right end fragments of bacteriophage X occurred withpartial denaturation within some of the molecules over a 40range. Table 1 shows the t., of the EcoRI fragments rDNA, AB,Xgt left arm, and Agt right arm (3), and their approximate basecompositions (4, 5). The t. is conceptually different from themelting temperature (tm) where i of the base pairs are nolonger paired.R-loop formation to saturationAn initial attempt to obtain a uniform population of DNA

FIG. 1. R-loops were made by heating 5 £g/ml of Agt-Sc11O9 DNA and 5 pg/ml total rRNA in 70% vol/vol formamide, 0.1M Pipes at pH 7.8,and 0.01 M Na3EDTA at 470 for 20 hr. The reaction was performed under oil in a sealed, siliconized glass tube. All 500RNA molecules examinedcontained an R-loop simila to those shown. The sample was mounted for electron microscopy by the formamide technique (2). Grids were stainedwith uranyl acetate and shadowed with Pt/Pd.

Biochemistry: Thomas et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 8

, 202

1

2296 Biochemistry: Thomas et al.

C00.0cr 0.5-

0

-cl

< ~~~~~~0z

0

02-

0.2

0 2 4 6 8 10

Time in hoursFIG. 2. R-loop formation was performed by heating 5 gg/ml

ofDNA and 5 pg/ml (0) or 0.5 Ag/ml (0) of total rRNA in 70% vol/volformamide, 0.1 M Pipes at pH 7.8, and 0.01 M Na&EDTA at 430. Thereaction was carried out under oil in a sealed siliconized glass tube.Samples were taken with 26-gauge thin-walled Teflon tubing. Thetubing was wiped with a Kimwipe soaked in ethanol to remove ad-hering oil. The sample was mounted for electron microscopy as de-scribed in Fig. 1.

molecules containing R-loops was made by heating a mixtureof Xgt-Scl109 DNA at 5 ,ug/ml and total rRNA at 5 sg/ml at470 for 20 hr. As shown in Fig. 1, clear R-loops were formed.Each of 500 DNA molecules examined contained an R-loop.Therefore, it is possible to saturate DNA for the formation ofR-loops. Measurements of the length of the DNA-RNA portionof the R-loops indicate that the rDNA fragment is homologousto about %4 of the length of the 26S rRNA. Single-stranded tailsof RNA were only observed on one end of each R-loop, indi-cating that the rDNA fragment is homologous to one end of the26S rRNA molecule.Determination of the rate of R-loop formation

The reaction conditions used here were selected to give a re-action rate which could be readily measured. Much higherreaction rates can be achieved by increasing the RNA and saltconcentrations. The course of R-loop formation under variousconditions was followed by examination of samples in theelectron microscope. The fraction ofDNA molecules containingan R-loop was scored. The size of the R-loop was not considered.A total of 100-200 molecules were scored at each time point and2-4 time points were taken for each condition. The accuracyof scoring in the electron microscope is greatest when i of themolecules have reacted. Therefore, time points near the i re-action time were scored and, in this work, the rate will be ex-pressed as the time for i reaction since this value is directlymeasured.

For the reaction DNA + RNA i DNA-RNA in which RNAis in molar excess so that its concentration does not changeduring the reaction, a plot of the log of the fraction of unreacted

0 0.2 \

a

0L50'' ' {E,450

zo 224700

C 4800

o0.2-U-

0..0 2 5 8 20 40 60

Time in hoursFIG. 3. R-loop formation was followed as described in the legend

to Fig. 2; an RNA concentration of 5 ,ug/ml was used. A number ofsamples were taken during each reaction. Those samples containingR-loops in approximately half of the DNA molecules were scored. Therepeat experiment at 430 shows a % reaction time similar to thatshown in Fig. 2.

DNA as a function of time should give a straight line. This issupported by the data in Fig. 2 at two different RNA concen-trations.

Rate of R-loop formation as a function of temperatureThe time course of R-loop formation was followed at a seriesof temperatures using a 20-fold molar excess of RNA. Some ofthese time courses are shown in Fig. 3. The i reaction time wasdetermined from such plots for each temperature. A plot of thelog of the i reaction time as a function of temperature is shownin Fig. 4. The rate of R-loop formation is very dependent uponthe temperature of the reaction. This plot also demonstrates thatthe temperature (tm) at which maximum reaction rate occursis very close to the tss (490) of the DNA. The temperature atwhich the largest and most uniform-appearing R-loops wereformed was from tss-20 to to +10. At t -90, most of the R-loops were quite small and were formed from small RNAfragments. At tss +20, denatured fragments of Xgt-Sc11O9 DNAwere observed.Rate of R-loop formation as a function of RNAconcentration and temperatureFor a simple bimolecular reaction, the rate of R-loop formationshould be directly proportional to the RNA concentration. Thisprediction is verified for the reaction at t. +10 as shown inTable 2. However, the rate at t .-60, as shown in Fig. 2 andTable 2, is increased by a factor of only 1.5 when the RNAconcentration is increased by a factor of 10. This behavior is alsofound at tss-90. Between tss-60 and t1, +10, the reaction rateconverts from one virtually independent of RNA concentrationto one directly proportional to RNA concentration.

Proc. Nad. Acad. Sci. USA 73 (1976)

Dow

nloa

ded

by g

uest

on

Aug

ust 8

, 202

1

Proc. Nati. Acad. Sci. USA 73 (1976) 2297

E10-

0

O.,-8 -4 0 +4 +8

Temperature relative to tss in. C

FIG. 4. The i reaction time was determined from the plots shownin Fig. 3. The t.l for duplex DNA (DNA.DNA) and for the RNA-DNAhybrid was determined by electron microscopy as described in Ma-terials and Methods.

Rate of R-loop formation as a function of ionicstrengthThe rate of DNA renaturation increases about 10-fold when thesodium ion concentration is increased from 0.17 M to 0.67 M(6). The rate of R-loop formation was determined for this sameincrease in salt concentration. The addition of 0.5 M NaCl tothe reaction mixture should increase the tm of DNA by 9° (6).The i reaction time at 560 (tss-20) in 0.5 M NaCl, 0.1I M Pipesat pH 7.8, 0.01 M Na3EDTA with 5 ,ug/ml ofDNA and 5 ,ug/mlof RNA is about 3.5 min. This represents approximately a15-fold increase in reaction rate over the same reaction in theabsence of NaCI.

Stability of R-loopsThe large R-loops used in this work, once formed, are quitestable. They are quantitatively retained after storage for severaldays at 5° in the formamide reaction mixture. Two samples,stored for 2 wk after dialysis into I M NaCI-0.01 M Tris-HClat pH 7.5-0.01 M NasEDTA or 0.1 M NaCl 0.01 M Tris HCIat pH 7.5-0.01 M NftEDTA, that originally contained 100%R-loops, were reduced to 99% and 98%, respectively. Thesedialyzed R-loop samples were also partially resistant to heat.About 25% of the R-loops remained in the sample containingI M NaCI after heating to 700 (tm-30°) for 5 min, and about

Table 2. Ratio of R-loop formation rate at 5 Aig/ml and0.5 jig/ml of total rRNA as a function of temperature

Temperature rela- Rate at 5 jig/mltive to tSS (0C) Rate at 0.5 gg/ml

-9 1.5-6 1.5-1 5.6+1 10

The rate of R-loop formation was determined as given in Fig. 2.

FIG. 5. A preparation of DNA containing 100% R-loops was di-alyzed into 0.1 M NaCl, 0.05 M Tris-HCl at pH 7.5, 10-4M EDTA.The cleavage reaction was started by the addition ofMgSO4 to 0.01M and ofEcoRI endonuclease. The mixture was digested at 370 for10 min. The resulting sample was mounted for electron microscopyas indicated in the legend to Fig. 1. The R-loop shown was formed witha fragment of 26S rRNA and therefore forms a smaller R-loop.

75% of the R-loops remained in the sample containing 0.1 MNaCl after heating to 50.' (tm- 300) for the same length oftime.

Since such stability of the R-loops was observed, their stabilityduring and after restriction endonuclease cleavage of R-loopscontaining DNA at 370 was tested. Fig. 5 shows one exampleof R-loops which were easily observed after cleavage of theDNA with EcoRI endonuclease.Removal of R-loopsUpon removal of formamide, R-loops are thermodynamicallyunstable at tm -300 and are displaced by branch migration (7).However, this displacement of large R-loops by identical DNAstrands is slow. Therefore, destruction of the RNA seems nec-essary for the elimination of R-loops. Two methods were suc-cessfully used. (1) The RNA was hydrolyzed by addition of 0.2M NaOH and incubated at 370 for 10 min. Denaturation of theDNA was prevented by addition of MgSO4 to a final concen-tration of 0.01 M prior to addition of alkali. After incubation,the solution was neutralized with Tris-HCI and the Mg ion wasremoved by increasing the EDTA concentrations to 0.02 M.Some DNA aggregation was observed in the electron micro-scope with this method. (2) The R-loops are quite sensitive toRNase. Incubation for 5 min at 370 in 0.1 M NaCl, 0.05 MTris-HCI at pH 7.5, 10-4 M EDTA with 10 Ag/ml of RNase Aresulted in the total loss of R-loops.

DISCUSSIONThe temperature for maximum rate (trm) of R-loop formationis within one degree of the t. of the DNA. From the values fort. of three XDNA EcoRI fragments with different known basecompositions (Table 1), the following equation relating ttnax tobase composition can be derived for the conditions used in theseexperiments:

tmax = 26.5 + 0.50 (%G+C).

If one assumes that the effect of salt concentration on the de-naturation of DNA in formamide is the same as in aqueous so-

Biochemistry: Thomas et al.

Dow

nloa

ded

by g

uest

on

Aug

ust 8

, 202

1

2298 Biochemistry: Thomas et al.

lution, then from Schildkraut's equation (8) in-70% formnamide:

tmax 39 + 0.50 (%G+C) + 16.6 log Na .

Also from Scbildkraut's equation, the effect of formamide canbe expressed as:

max = 815 + 0.50 (%G+C) + 16.6 log Na*

-0.60 (% formamide).

These equations have not been rigorously tested and may notapply if the conditions are significantly varied from those usedin this work.

R-loops are not thermodynamically stable at tm -300 afterthe formamide is removed since the RNA is slowly displacedby the identical DNA strand through branch migration (7). Atlow temperatures (tm -700), in the absence of formamide, thestability may result from random base pairing in the single-stranded regions which block branch migration. At highertemperatures (tm -300), branch migration commences and theRNA is found to be partially displaced. However, if displace-ment occurs at both ends, the RNA must thread through theR-loop once for each turn of duplex RNA-DNA hybrid re-moved. As theRNA is displaced, the loop becomes progressivelysmaller, and, concurrently, the displacement may becomeprogressively slower. Consistent with this notion are the ob-servations that, after heating in the absence of formamide, mostR-loops are very small and that the rate of displacement at tm-300 is slower in low ionic strength buffer where electrostaticrepulsion between the strands is greater. As evidenced by theRNase sensitivity of the R-loops, if the RNA is destroyed as itis displaced, these blocks to the branch migration are removed.The formation of R-loops is probably equally complicated.

The most likely mechanism for their formation is the generationof denatured regions in the duplex DNA and nucleation of RNAwithin these denatured regions. The size and frequency of thesedenatured regions would depend upon the temperature, beingsmall and infrequent below the tss of the DNA. If the size of adenatured region in theDNA is small relative to the RNA, thenit seems unlikely that irreversible nucleation can occur easilyin the center of the RNA because the RNA would have to threadthrough a very small loop. Nucleation in this case probablyoccurs most readily at the end of the RNA. Thus, if there arenonhomologous sequences on the end of the RNA, as is the casein this study, the rate of R-loop formation may be reduced.

The observation that the rate of R-loop formation is virtuallyindependent of the RNA concentration at temperatures belowt, -50 is difficult to understand. One possible explanation isthat the rate-limiting step under these conditions is the for-mation of denatured regions. Another possible explanation isthat the rate-limiting step under these conditions is subsequentto the nucleation event. The addition of base pairs to the nu-cleation complex may be slow due to steric interference inwinding. The nucleation complexes that do not proceed to fullR-loop formation dissociate before or during mounting for

electron microscopy. The rate limiting sep above tss should bethe rate of RNA nucleation. Therefore, above t, the rate ofR-loop formation should be directly proportional to the RNAconcentration as is found in this study. Also, as is found in thisstudy, the rate of R-loop formation above the tss of the DNAshould decrease because of the decreased stability of the nu-cleation events. With low RNA concentrations at temperaturesbelow tag, the rate-limiting step could be the RNA nucleation.Consistent with this suggestion is the observation that, at lowRNA concentration between t. -60 and t., the rate of R-loopformation is fairly independent of temperature.

It is not clear if very small R-loops can be generated since ourinitial attempts to form R-loops with yeast 5S RNA have notbeen successful. Small R-loops have an additional complicationin that the RNA is probably rapidly displaced.The experiments presented here indicate that the R-loop

formation reaction is quite complex. There are additionalvariables that were not quantitatively investigated. For ex-ample, the secondary structure in the RNA which is not totallyremoved under these conditions could affect the R-loop for-mation rate. Also, the size of theRNA may well affect the ratesince it was observed that the ratio of R-loops derived fromsmall RNA to those derived from large RNA was greater attemperatures below t. than at temperatures above tss.The base sequence and composition may also affect the rate

of formation and stability of R-loops. R-loops are sufficientlystable to withstand dialysis into a buffer of low ionic strength(0.05 M Tris.HCI at pH 7.5, 0.1 M NaCl) and cleavage with anEcoRI restriction endonuclease. This should greatly facilitatethe mapping of a number of rRNA and mRNA sequences inDNA. The end of the R-loop can be rather precisely mappedrelative to nearby restriction sites.The use of R-loops should facilitate the physical isolation of

specific genes. For example, R-loops might be formed on un-fractionated DNA with use of a specific mRNA. The resultingpreparation containing duplex DNA and R-loops on the specificgene sequence could be sedimented to equilibrium in a CsCldensity gradient. The R-loop should give the DNA containingthe specific gene sequence a density shift; this would result inconsiderable purification.

1. Robberson, D. L., Kasmatsu, H. & Vinograd, S. (1972) Proc. Natl.Acad. Sd. USA 69,737-741.

2. Davis, R. W., Simon, M. N. & Davidson, N. (1971) in Methods inEnzymology, eds. Grossman, L. & Moldave, K. (Academic Press,New York), Vol. 21, pp. 413-428.

3. Thomas, M., Cameron, J. R. & Davis, R. W. (1974) Proc. Natl.Acad. Sci. USA 71,4579-4583.

4. Davidson, N. & Szybalski, W. (1971) in The BacteriophageLambda, ed. Hershey, A. D. (Cold Spring Harbor Laboratories,Cold Spring Harbor, N.Y.), p. 65..

5. Retel, J. & Van Keulen, H. (1975) Eur. J. Biochem. 58,51-57.6. Wetmur, J. G. & Davidson, N. (1968) J. Mol. Biol. 31,349-370.7. Lee, C, S., Davis, R. W. & Davidson, N. (1970) J. Mol. Biol. 48,

1-22.8. Schildkraut, C. & Lifson, S. (1965) Blopolymers 3, 195.

Proc. Natl. Acad. Sci. USA 73 (1976)

Dow

nloa

ded

by g

uest

on

Aug

ust 8

, 202

1