Interaction of linear homologous DNA duplexes via Holliday junction formation

Eur. J. Biochem. 268, 7±14 (2001) q FEBS 2001

P R I O R I T Y PA P E R

Interaction of linear homologous DNA duplexes via Holliday junctionformation

Marianna G. Yakubovskaya1, Anna A. Neschastnova1, Karen E. Humphrey2, Jeff J. Babon2, Vladimir I. Popenko3,Margaret J. Smith4 Andrianna Lambrinakos2, Zhanna V. Lipatova1, Marina A. Dobrovolskaia1, Roberto Cappai4,Colin L. Masters4, Gennady A. Belitsky1 and Richard G. Cotton2

1Carcinogenesis Institute, Cancer Research Centre, Russian Academy of Medical Sciences, Moscow, Russia; 2Mutation Research Centre,

St. Vincent's Hospital, Fitzroy, Victoria, Australia; 3Engelgardt Institute of Molecular Biology, Russian Academy of Sciences, Moscow, Russia;4Department of Pathology, University of Melbourne, Parkville, Victoria, Australia

Interaction of linear homologous DNA duplexes by formation of Holliday junctions was revealed by

electrophoresis and confirmed by electron microscopy. The phenomenon was demonstrated using a model of five

purified PCR products of different size and sequence. The double-stranded structure of interacting DNA

fragments was confirmed using several consecutive purifications, S1-nuclease analysis, and electron microscopy.

Formation of Holliday junctions depends on DNA concentration. A thermodynamic equilibrium between

duplexes and Holliday junctions was shown. We propose that homologous duplex interaction is initiated by

nucleation of several dissociated terminal base pairs of two fragments. This process is followed by branch

migration creating a population of Holliday junctions with the branch point at different sites. Finally, Holliday

junctions are resolved via branch migration to new or previously existing duplexes. The phenomenon is a new

property of DNA. This type of DNA±DNA interaction may contribute to the process of Holliday junction

formation in vivo controlled by DNA conformation and DNA±protein interactions. It is of practical significance

for optimization of different PCR-based methods of gene analysis, especially those involving heteroduplex

formation.

Keywords: branch migration; DNA association; DNA complex; duplex interaction; Holliday junction.

The high non-specific background of mutation detectionanalysis prompted an investigation of the possible reasons fornon-specific DNA cleavage. Modern methods for the detectionof random mutations are predominantly PCR-based, with manyof them dependent on recognizing mismatches in hetero-duplexes consisting of wild-type and mutant strands [1±3]. Oneof these techniques, chemical cleavage of mismatches (CCM),is based on chemical modification and cleavage of mismatchedC or T residues at the position of mutation followed bydenaturing gel electrophoresis for identification of the cleavage[4±6]. Another approach for mutation detection in hetero-duplexes is enzyme mismatch cleavage (ECM) using repairenzyme [7] and resolvase enzymes: T4 endonuclease VII [8,9]and T7 endonuclease I [10]. As both CCM and ECM are basedon the linear structure of heteroduplexes, we proposed thathigh background cleavage may be explained by formation ofnon-linear branched DNA structures cleaved at the cross-points.

Purified duplex interaction without protein participation wasrevealed for DNA fragments containing CA repeats [11], andsome effect of hydrophobic surface on DNA±DNA interactionwas shown [12±14]. DNA±DNA interaction was explained byduplex±duplex interaction without strand exchange [11].

Belotserkovskii & Johnston [12,14] proposed extended duplexdenaturation followed by random association of ssDNAfragments resulting in DNA complex formation with randomHolliday junctions. Recently, we demonstrated the phenomenonof linear duplex interaction for the PCR product of p53 cDNA[15], and Gaillard et al. [16] described formation of low-mobility structures from two non-repetitive 190-bp fragmentsrestricted from the genome of simian virus 40 (SV40). Weproposed that DNA±DNA interaction via Holliday junctionformation is initiated by nucleation of several dissociated basepairs at the ends of DNA fragments [15]. Our hypothesis isconsistent with previous work of Panyutin & Hsieh [17]demonstrating interaction of two homologous duplex DNAseach having two single-stranded tails at one end that arecomplementary to the corresponding single-stranded tails of theother duplex. On rapid annealing of complementary single-stranded tails of the duplexes, a Holliday junction was formedand branch migration resulted in complete strand exchange andformation of two duplex products. In our experiments,interaction of completely homologous duplexes is consideredand it is proposed to occur because of the constant process ofdissociation and association of several base pairs at the veryends of DNA fragments.

Four-way complex DNA structures required for the processof recombination were proposed by Holliday [18], and theirformation in vivo was confirmed by electron microscopy ofDNA structures of twice the size of plasmids, phages, andmitochondrial DNAs [19,20]. A set of proteins was shown toparticipate in Holliday junction formation altering the topologyof the DNA [21] enabling single-strand break appearance and

Correspondence to M. G. Yakubovskaya, Carcinogenesis Institute,

Cancer Research Centre, Russian Academy of Medical Sciences,

Kashirskoye Shosse 24, Moscow, 115478 Russia.

Fax:/Tel.: 7 095 323 58 22, E-mail: [email protected]

Abbreviations: CCM, chemical cleavage of mismatches; ECM, enzyme

mismatch cleavage.

(Received 5 September 2000; accepted 30 October 2000)

approach of homologous DNA regions [22,23]. SyntheticHolliday junctions formed by half-complementary oligonucleo-tides have been shown to be cleaved both chemically [24,25]and by a number of enzymes [26,27]. It is in accordance withour proposal that background cleavage during CCM or EMCmay be caused by the appearance of non-linear DNA structuresin homologous duplex solution.

The goal of this study was to demonstrate DNA±DNAinteraction without protein participation, using the model ofpurified homologous DNA duplexes of different size andsequence, and to elucidate the mechanism further.

M A T E R I A L S A N D M E T H O D S

PCR products

Amplification was carried out using Gibco-BRL Life Tech-nologies or Biomaster Taq DNA polymerase in reactionsperformed with PTC-100TM (MJ Research Inc., San Francisco,CA, USA) or Corbett Research (Corbett, Sydney, Australia)thermocyclers. For the amplification of DNA fragments, weused the primers and protocols as follows: for the 1249-bpfragment of p53 cDNA: forward: 5 0-TGCTTTCCACGACGGT-GACACGCTTCCCTGGATTG-3 0, reverse: 5 0-TGGAGAATG-TCAGTCTGAGTCAGGCCC-3 0, 95 8C for 2 min; 30 cycles of95 8C for 1 min, 58 8C for 1 min, 72 8C for 1 min; 72 8C for10 min; 4 8C; for the 1022-bp PS1 gene: forward: 5 0-ATA-CCTAATCTGGAGCCTG-3 0, reverse: 5 0-CACACCATTGTT-GAGGAGT-3 0, 95 8C for 2 min; 30 cycles of 95 8C for 1 min,55 8C for 1 min, 72 8C for 1 min; 72 8C for 10 min; 4 8C; forthe 550-bp fragment of the mouse b-globin gene promoterregion: forward: 5 0-GCACGCGCTGGACGCGCAT-3 0, reverse:5 0-AGGTGCCCTTGAGGCTGTCC-3 0, 95 8C for 2 min; 30cycles of 95 8C for 30 s, 60 8C for 45 s , 72 8C for 1 min; 72 8Cfor 10 min; 4 8C. Plasmids with p53 cDNA, PS1 fragment andmouse b-globin gene promoter region were kindly given byP. Chumakov [28], the Neurogenetics and NeurotransgenesisGroups, Mayo Clinic, Jacksonville, FL, USA [29] andR. Myers, respectively. The 209-bp fragment of p53 exon 8was amplified using genomic DNA from cell lines MDA-231,DU-145 and PC-3, using the forward and reverse primers5 0-ACTGCCTCTTGCTTCTCTTTTC-3 0 and 5 0-TTGGTCTC-CTCCACCCGATC-3 0 respectively and the protocol 95 8C for3 min, 30 cycles of 95 8C for 1 min/58 8C for 1 min/72 8C for1 min, 72 8C for 10 min, 4 8C. The 118-bp fragment of theDras1 gene was amplified from Drosophila melanogastergenomic DNA, using the forward and reverse primers5 0-TGACGGAATACAAACTGGTCG-3 0 and 5 0-TAAGAGTC-CTCGATTGTGGGG-3 0 respectively and the protocol 95 8Cfor 2 min, 30 cycles of 95 8C for 1 min/54 8C for 1 min/72 8Cfor 1 min, 72 8C for 10 min, 4 8C. Genomic DNA was preparedby the standard phenol/chloroform DNA purification protocol[30].

Gel electrophoresis (non-denaturing)

Native 10% polyacrylamide gels were run at room temperatureat 100 V in 89 mm Tris/borate buffer, pH 7.5, containing 2 mmEDTA (Tris/borate/EDTA), 1.2% agarose gels were run at roomtemperature at 60 V in 89 mm Tris/acetate buffer, pH 7.5,containing 2 mm EDTA (Tris/acetate/EDTA) or Tris/borate/EDTA, or in 89 mm Tris/acetate buffer, pH 7.5, with 10 mmMgCl2 (Tris/acetate/MgCl2), when MgCl2 influence wasestimated.

Gel electrophoresis (denaturing)

The ABI 377 DNA Sequencer was used with 5% denaturingpolyacrylamide gel prepared by a standard protocol. genescansoftware (ABI/PE, Melbourne, Australia) was used to analysethe fluorescently labelled fragments. Also, purified PCRproducts were subjected to electrophoresis in 5% or 6%denaturing polyacrylamide gels, transferred to positivelycharged nylon membrane (Sigma) and hybridized with one ofthe primers 5 0-labelled using polynucleotide kinase and[g-32P]ATP [31].

DNA purification from agarose gels

The gel slice with the ds PCR product or DNA structure ofinterest was excised from agarose gel after electrophoresis andpurified using BresaClean PCR product purification kit(Bresatec, Adelaide, Australia) following the manufacturer'sinstructions, or using DNA-binding silica matrix. The gel slicewas incubated in 2 vol. 6 m NaI at 55 8C for 30 min (until fullmelting), then the appropriate volume of silica matrix wasadded and the mixture was incubated at 24 8C for 30 min. Aftercentrifugation (2000 g for 2 min) the pellet was washed twicewith washing buffer [50 mm NaCl, 10 mm Tris/HCl, pH 7.5,2.5 mm EDTA, 50% (v/v) ethanol] and then DNA was eluted in10 mm Tris/HCl buffer, containing 1 mm EDTA (Tris/HCl/EDTA) or 10 mm Tris/HCl buffer, containing 50 mm NaCl and1 mm EDTA (Tris/HCl/NaCl/EDTA) at 55 8C for 30 min.

Preparation of DNA-binding matrix

First 10 g silica (Sigma S-5631) was incubated in 100 mLphosphate-buffered saline, pH 7.4, containing 137 mm NaCl,2.7 mm KCl, 4.3 mm Na2HPO4 and 1.47 mm KH2PO4 for 2 h.The solution was centrifuged and the supernatant removed. Theprocedure was repeated and the pellet resuspended in 100 mL3 m NaI.

DNA purification from polyacrylamide gels

The slice was excised from polyacrylamide gel, crushed andleft overnight in a solution containing 150 mm NaCl, 50 mmTris, and 20 mm EDTA (pH 8). The solution was thencentrifuged (14 000 g for 10 min), and DNA was ethanol-precipitated from the supernatant.

DNA purification by electroelution and filtration

After agarose gel electrophoresis for 40 min, the gel slice underthe DNA band of interest was excised and the resulting wellwas filled with electrophoresis buffer. Electrophoresis wascontinued for 2±3 min, then DNA-containing buffer wasremoved from the well and DNA was filtered on UltrafreeMicrocentrifuge Filters (Sigma M-1161).

Radioactive labelling

Polynucleotide kinase (Progen, Darra, Queensland, Australia;10 U´mL) and [g-32P]ATP (NEN; 3000 Ci´mmol21, 10 mCi´mL21)was used for 5 0-labelling of the 1249-bp p53 PCR product orprimers [30,31].

S1-nuclease analysis

The amount of S1-nuclease needed to remove the ssDNA butleave the dsDNA was determined by titration using 10, 5 and

8 M. G. Yakubovskaya et al. (Eur. J. Biochem. 268) q FEBS 2001

2 U S1-nuclease (Boehringer-Mannheim GmbH). The PCRproduct was then incubated with 5 U S1-nuclease in standardbuffer for 10 min at 37 8C, to remove the ssDNA. PCR productwas then purified using DNA-binding matrix.

Electron microscopy

DNA for electron microscopy was prepared using the proteinmonolayer technique [15,32]. The specimens were investigatedin JEM 100CX electron microscope (JEOL) at 80 kV.

R E S U LT S A N D D I S C U S S I O N

Appearance of tetramers in purified PCR product solutions

Agarose gel electrophoresis with visualization of DNA byethidium bromide staining was used to investigate solutions ofpurified PCR products of five different genes: (a) the 1249-bpPCR product of p53 cDNA, exons 2±9; (b) the 1022-bp PCRproduct of PS1 cDNA; (c) the 546-bp PCR product includingmouse b-globin gene promoter region; (d) the 209-bp PCRproduct of p53 gene, exon 8; (e) the 118-bp PCR product of theDras1 gene of D. melanogaster. PCR products were purifiedusing BresaClean kit (Bresatec) or DNA-binding silica matrix.After purification of PCR products from the corresponding gelslice followed by electrophoresis, small amounts of DNAstructures about twice the size of duplex DNA and of highmolecular mass appeared (Fig. 1A). This was observed whengel was stained with ethidium bromide both before and afterelectrophoresis. The quality of the purification of the PCRproducts was assessed by Southern-blot hybridization withradiolabelled primers after denaturing gel electrophoresis:

ssDNA fragments only of the expected size of the PCR productwere detected. To determine the apparent size of DNA formingthe structures with lower mobility than the ds PCR product, thePS1 gene fragment was amplified using fluorescently labelledprimers. After agarose gel electrophoresis, the bands corre-sponding to the ds PCR product, DNA structures of twice themolecular mass of the ds PCR product and high-molecular-mass DNA complexes were excised from the gel. DNA wasextracted and then subjected to denaturing gel electrophoresisin an ABI 377 DNA sequencer for 12 h. The electrophoreto-grams of DNA from each band excised had only one peakcorresponding to the expected size of the PCR product.

Although more intensely stained bands were observed forlonger PCR products, the pattern of the additional bands afterpurification and electrophoresis was similar for the five PCRproducts studied. Therefore it was proposed that the appearanceof higher-molecular-mass DNA structures was not due toconformation changes in the secondary structure of theduplexes but to the formation of DNA complexes from twoor more ds PCR products therefore with four or more DNAstrands. The term `tetramer' is used to indicate DNA ofapparent size twice that of the ds PCR product [33].

Tetramer formation in different conditions

We proposed that tetramer formation was the consequence ofheating during purification, as the protocol requires incubationsat 55 8C (BresaClean kit). We did not notice the appearance ofadditional bands immediately after PCR. This may be explainedby the use of optimal PCR conditions for the amplification untilthe last stage and by the short incubation at 72 8C at the veryend when the Taq DNA polymerase elongates DNA to perfectduplex [2]. To test whether incubation of PCR product solutionsat 55 8C for 4 h causes the appearance of tetramers, weanalysed PCR solutions either directly or after concentration byethanol precipitation. The appearance of tetramers after theincubation was confirmed by electrophoresis for all PCRfragments (Fig. 1B). However, after the incubation of PCRproducts, a certain amount of ssDNA appeared which was notobserved after duplex purification. This appearance of ssDNAafter incubation of the homologous duplex solution agreed withthe previous observation of [12]. When the 118-bp Dras1 dsPCR product was purified from a polyacrylamide gel and thenDNA was concentrated by ethanol precipitation from thesupernatant, no additional bands were detected. However,incubation of the purified Dras1 ds PCR product solution(20 mg´mL21 in Tris/HCl/NaCl/EDTA) at 608C for 4 h led totetramer formation (Fig. 1B). When the DNA purificationprocedure was carried out at 37 8C, which is much lower thanthe temperature recommended in the BresaClean kit protocol,this did not prevent the formation of tetramers, but theadditional high-molecular-mass band was less intense.

When the PCR products were electrophoresed in Tris/acetate/MgCl2 with 100 mm MgCl2, a significant portion ofDNA remained in the wells and did not enter the agarose gel;however, the additional band of tetramers was still seen. Itcorrelates with the results obtained by differential scanningcalorimetry and optical densitometry showing that Mg21 ionsinitiated DNA aggregation in concentrated DNA solutions [34].

From these results, we can conclude that the tetrameric high-molecular-mass DNA structures appear in solutions of homo-logous DNA fragments when different temperatures, reactionbuffers (Tris/HCl/EDTA, Tris/NaCl/HCl/EDTA, PCR solution)and electrophoresis buffers (Tris/borate/EDTA, Tris/acetate/EDTA, Tris/acetate/MgCl2) are used and that the ratio of the

Fig. 1. Agarose gel electrophoresis. (A) M1, fX174/HaeIII; M2,

l/HindIII, M3, pBR 322/TaqI. Before purification: lane 1, the 1249-bp

PCR product of the p53 gene. After purification using DNA-binding matrix

and incubation at 55 8C for agarose melting: lane 2, the 1249-bp PCR

product of the p53 gene; lane 3, the 1022-bp PCR product of the PS1 gene;

lane 4, the 550-bp PCR product of the mouse b-globin gene; lane 5, the

209-bp PCR product of the p53 gene. (B) Lane 1, the 1022-bp PCR product

of the PS1 gene directly after amplification. After incubation at 55 8C for

4 h: lane 2, the 1022-bp PCR product of the PS1 gene; lane 3, the 550-bp

PCR product of the mouse b -globin gene; lane 4, the 209-bp PCR product

of exon 8 of the p53 gene; lane 5, the 118-bp PCR product of the Dras1

gene after purification from polyacrylamide gel; lane 6, the purified 118-bp

PCR product of the Dras1 gene after incubation at 60 8C for 4 h; lane 7,

purification of the 1249-bp PCR product of the p53 gene using extraction of

DNA from agarose gel slices using 37 8C for melting of the agarose slice;

lane 8, purification and electrophoresis of the 1249-bp PCR product of p53

in a Tris/acetate/MgCl2 with 100 mm MgCl2.

q FEBS 2001 Interaction of linear homologous DNA duplexes (Eur. J. Biochem. 268) 9

intensity of the additional bands may be influenced by thesefactors.

Tetramer formation was shown to be strongly dependent onDNA concentration. Moreover, when purified PCR productsolutions of concentration 100 mg´mL21 and 200 mg´mL21

were analysed, a number of DNA structures corresponding toconsecutive increases in molecular mass were identified(Fig. 2).

Double-stranded structure of interacting DNA fragments

To confirm that tetramers were formed by DNA duplexes,S1-nuclease analysis was performed. Before purification, thePCR product was subjected to S1-nuclease treatment to destroy

ssDNA and to leave only perfect duplexes in the solution.S1-nuclease activity was titrated and sufficient was taken forfull digestion of ssDNA, obtained by denaturing of the portionof dsDNA used in the experiments. The efficiency ofS1-nuclease treatment was also confirmed by disappearanceof the bands corresponding to the primers. The duplexeswere then purified using agarose gel electrophoresis andDNA-binding matrix. The band of tetramers was detectedafter duplex purification from PCR product previously treatedwith S1-nuclease (Fig. 3). Therefore, the removal of ssDNAfrom PCR-product solution did not prevent formation of thetetramers. This result is in accordance with data obtained byGaillard et al. [16] using the model of poly(CA) containing190-bp fragments restricted from the genome of SV40. Theyshowed the appearance of DNA complexes of low mobility,which harboured all perfect duplex structures except for a veryshort ssDNA region linked to the poly(CA) sequence.

When we performed three consecutive purifications of theduplexes, additional bands of equal intensity appeared after thefirst, second, and third purification (Fig. 3). This confirms thatthe tetramer formation was not associated with amplificationerrors. If contamination had played any part in tetramerformation, the intensity of the additional bands would havebeen decreased after the second and third purifications.

Tetramer structure

Further investigations were carried out to elucidate the structureof the tetramers. Two types of complex DNA structures formedby four DNA strands have been described: Holliday junctionsand G-structures [17,35]. G-structures can be formed by G-richsequences, but on addition of the Watson±Crick complemen-tary strands, four-stranded structures dissociate quickly to formDNA duplexes [36]. In contrast, thermodynamic equilibriumbetween the synthetic Holliday junctions and perfect duplexeswas confirmed using eight synthetic oligonucleotides, senseand antisense sets of four half-complementary oligonucleotides,forming correspondingly four duplexes or two Hollidayjunctions [33]. Electron-microscopic analysis was used toconfirm the existence of the Holliday junctions in vivo [19,20].We tried to extract tetramers for electron microscopy. When thetetramer band or high-molecular-mass DNA band was excisedfrom the agarose gel and DNA was extracted using DNA-binding matrix and subjected to electrophoresis, the mostintense band on the gel corresponded to the band of ds PCRproduct. However, the additional bands of higher molecularmass were also present. The ratio of the intensity of these bandsafter purification of DNA from additional bands andre-electrophoresis was very similar to that observed initiallyin the solution of the purified duplexes. These results show that

Fig. 2. Electrophoresis of the 1249-bp PCR product of p53 cDNA

solutions with different DNA concentrations. M1, fX174/HaeIII; lane 1,

10 mg´mL21; lane 2, 25 mg´mL21; lane 3, 50 mg´mL21; lane 4,

100 mg´mL21; lane 5, 200 mg´mL21.

Fig. 3. S1-nuclease treatment of PCR mixture and the results of PCR-

product purification. (A) Lane 1, unpurified 1249-bp PCR product of p53

cDNA; lanes 2±4, unpurified PCR product (the same amount as in lane 1)

of 1249 bp of p53 cDNA treated for 10 min at 37 8C with 10, 5, 2 units of

S1-nuclease respectively; lanes 5±7, unpurified PCR product (the same

amount as in lane 1) of 1249 bp of p53 cDNA, denatured and then treated

for 10 min at 37 8C with 10, 5, 2 units of S1-nuclease respectively; lane 8,

unpurified PCR product (the same amount as in lane 1) of 1249 bp of p53

cDNA, denatured; lane 9, the purified 1249-bp PCR product of p53 cDNA,

25 mg´mL21; lane 10, the purified 1249-bp PCR product of p53 cDNA

treated before purification with 5 U S1-nuclease for 10 min at 37 8C,

25 mg´mL21. (B) Agarose gel electrophoresis of DNA purified from

agarose gel using DNA-binding matrix (lanes 1±3) or electroelution and

filtration (lanes 4 and 5). Lane 1, the 1249-bp PCR product of p53 cDNA

after three consecutive purifications; lane 2, the purified 1249-bp PCR

product of the p53 gene; lane 3, the purified DNA from the gel slice of the

tetramers; lane 4, the purified 1249-bp PCR product of the p53 gene; lane 5,

the purified DNA from the band of tetramers.

Fig. 4. Electron-microscopic analysis of the solutions of (A) the

tetramers, (B) the ds PCR product of p53, and (C) trans-isomers and

cis-isomers of Holliday junctions with different positions of the

cross-point.


there is an equilibrium between the duplexes and higher-molecular-mass DNA structures (Fig. 3). At first, because ofthis equilibrium, direct experimentation on the additional DNAstructures was difficult because tetramers mainly convertedback to apparent duplexes during purification. We took intoconsideration the fact that the period of tetramer conversion toduplexes corresponded to that reported by Panyutin & Hsieh[17]. These authors described Holliday junction formation bytwo homologous duplex DNAs each having two single-strandedtails at one end that are complementary to the correspondingsingle-stranded tails of the other duplex. Complete dissociationof Holliday junction occurred in their system over 60±100 minbecause of branch migration through a 956-bp duplex. Wethought that, in our experiments, branch migration may haveoccurred during tetramer purification resulting in reverseduplex formation from four-stranded DNA complexes. Toenrich the DNA solution with tetramers, we decided to speed upthe DNA purification. By using electroelution and filtration onmicrocentrifuge columns, we purified DNA more rapidly at alower temperature and in the absence of any chaotropic agent.As a result, tetramers were trapped (Fig. 3). The solution oftetramers was analysed by electron microscopy. As shown inFig. 4A, it contained a high proportion of x structures. In thesolution of duplexes purified by electroelution and filtration,the band pattern was the same as after purification using DNA-binding matrix (Fig. 3), and electron-microscopic analysisrevealed mainly linear DNA fragments of the correct size(Fig. 4B).

These x structures could not be the result of overlapping asthe concentration of DNA was low: the DNA structures weresmaller than the distance between them, and the use of lowerDNA concentrations prepared directly before electron micro-scopy did not change the ratio between x structures and linearfragments. Electron microscopy of DNA solution extractedfrom the band of the ds PCR product revealed predominantlylinear fragments, whereas many x structures were detected inthe analysis of DNA solution obtained from the tetramer band.This occurred despite the fact that the DNA concentration wasthe same in the two DNA solutions. All four-stranded DNAstructures observed in the tetramer fraction were symmetrical,whereas the point of branch migration was located at differentpositions and both cis-conformation and trans-conformationstructures were identified (Fig. 4C). The symmetry of all the xstructures observed also confirmed that it was not overlapping.Our data on the different locations of the point of branchmigration fit the theory that Holliday junctions cannot beviewed as stable structures but rather as an equilibrium mixtureof conformational isomers [25,37,38] and that natural Hollidayjunctions with homologous sequence symmetry branch migrate

to create a population of DNA structures that have the branchpoint at different sites.

Proposed mechanism of homologous DNA fragmentinteraction

According to the modern theory of DNA duplex structure[39] and the analysis of duplex melting and fraying [40],Watson±Crick interaction of the terminal base pairs of linearDNA duplexes are less stable than in the internal regions, whichis related to the decrease in stacking interaction at fragmentends [39]. Correspondingly, we propose that DNA±DNAinteraction occurs via nucleation of several dissociated terminalbase pairs of two fragments, followed by branch migrationcreating a population of Holliday junctions with the branchpoint at different sites, which are resolved to new or previouslyexisting DNA duplexes (Fig. 5). Our hypothesis is consistentwith the work of Panyutin & Hsieh [17] showing interaction oftwo homologous duplex DNAs each having two single-strandedtails at one end that are complementary to the correspondingsingle-stranded tails of the other duplex. Rapid annealing of thesingle-stranded tails of two duplexes results in Hollidayjunction formation at one end of the duplexes. Branchmigration to the opposite end of the duplexes leads to completestrand exchange and formation of new duplexes. In ourexperiments interaction of absolutely homologous duplexes isconsidered and it is proposed to occur analogously to theprocess described by Panyutin and Hsieh because of theconstant process of dissociation and association of several basepairs at the ends of DNA duplexes.

Our proposal is confirmed by the fact that we did not ever seetwo cross-points in any complex DNA structure formed by twoPCR products; these would have been observed in cases ofinteraction of internal regions of dsDNA fragments. During ourDNA purification procedure, we did not observe any ssDNAformation, only tetramers and structures of high molecular massbeing identified. Therefore, in our experiments, tetramerformation could not have been related to DNA duplexdissociation followed by reassociation of ssDNA.

Gaillard et al. [16] showed strand exchanging when duplexeswere incubated in the presence of one of the duplex strands.This was confirmed by Renalldo et al. [41], and analysis of thekinetics of oligonucleotide replacement revealed accordance ofthe strand-exchange process with the sequential-displacementmechanism. We believe that homologous duplex interactionshould also be considered a sequential-displacement processbecause it happens via nucleation of terminal base pairsfollowed by branch migration during which new duplexes areformed.

It is interesting to compare our hypothesis with the results ofGaillard et al. [16]. Circled DNA duplexes were shown not tobe able to interact, which confirms the necessity for duplexends. Also, Gaillard et al. showed the formation of low-mobility structures by linear duplexes closed by hairpins. Weexplain such an interaction by the fact that the region at theends of hairpins is single-stranded [39]. Interaction of thesesingle-stranded hairpin ends may be analogous to nucleation ofdissociated ends of duplexes.

The structure of DNA duplexes with constantly dissociating±associating bases at the ends described by Saenger [39] couldprovide another explanation for why hydrophobic surfaces caninteract with DNA and stimulate DNA±DNA interaction[12,14]. As ssDNA interacts more efficiently with hydrophobicsurfaces, it is possible that several dissociated base pairs at theends of DNA fragments interact with polypropylene, leading to

Fig. 5. Proposed mechanism of DNA±DNA interaction. Step 1, DNA

duplexes in non-denaturing conditions; step 2, interaction of two duplexes

at two or three dissociated nucleotides at the ends of the fragments; 3,

formation of a Holliday junction; 4, branch migration in the formed

Holliday junction; 5, resolving of the Holliday junction with new duplex

formation.


an increase in local DNA concentration and stabilization ofduplex ends in the dissociated form. The effect of polypro-pylene on DNA±DNA interaction also needs to be analysed,but we believe that the mechanism of DNA complex formationvia dissociation and random reassociation proposed byBelotserkovskii & Johnston [14] for CA(30)-containing DNAfragments cannot be applied to fragments of non-repetitivesequences. When duplex purification was performed byelectroelution and filtration, DNA duplex solution could onlyhave been in contact with polypropylene during centrifugationon columns, and when this time was decreased to 7±10 min,tetramer formation was still very intense (Fig. 3, lane 5). In allour experiments involving duplex purification, we observedtetramer formation, but never noticed the appearance of ssDNA.Moreover in all DNA solutions analysed by electron micro-scopy, only perfect duplexes or x structures formed by dsDNAwithout any single-stranded regions were observed.

Significance of the phenomenon

DNA±DNA interaction in purified PCR product solutionsrevealed the ability of native homologous duplexes to formHolliday junctions. It was previously thought that linear DNAduplexes could not interact without protein participation, andnative Holliday junctions in purified DNA solutions wereidentified only in supercoiled circular DNA duplexes harbour-ing palindromic sequences. Therefore, our finding providesanother model system for the investigation of native Hollidayjunctions, which are the main intermediates in the process ofrecombination.

It is not yet clear whether the ability of DNA linear duplexesto interact without protein participation is important inrecombination because the system described does not replicatethe conditions in vivo. However, it is logical that the well-ordered process of DNA duplex interaction contributes to theprocess controlled by DNA conformation and DNA±proteininteraction.

Taking together our results on homologous duplex inter-action, the well-known dependence of branch migration on thesequence of Holliday junction cross-points, and the results ofGaillard & Strauss [11] showing the location of the cross-pointin DNA complexes formed by CA(30)-containing linearplasmid in the region of CA repeats, the biological role ofrepeats in the eukaryotic genome as `hot-spots' for geneticrecombination [42] could have a molecular explanation. Thespeed of branch migration may decrease in the region ofrepeats, and the probability of Holliday junctions resolving inthese regions becomes correspondingly higher. However, thisproposal and the results of Gaillard et al. [11,16] are incontradiction with the results of Benet & Azorin [43], whoshowed similar branch migration characteristics for non-repetitive fragments and fragments containing (CA) repeats incitric/phosphate buffer at different pHs. Furthermore, they didnot observe any DNA complexes in solutions of CA(30)-containing fragments. This may be connected to the absence ofmagnesium in the DNA solutions and the correspondingly veryhigh speed of branch migration [17]. Strong inhibition ofbranch migration was demonstrated by Benet & Azorin [43]only for CT-containing fragments at low pH, when triplexformation was observed [44].

Homologous duplex interaction to form Holliday junctions isone more piece of evidence in support of the hypothesis ofGierer [45], who proposed that DNA may form branchingstructures analogously to tRNA. This hypothesis was confirmed

previously by the observation of a DNA cruciform in dsDNAharbouring palindromic sequences.

The ability of homologous duplexes to form branchedstructures is of practical significance in different PCR-basedanalyses, in particular those involving a heteroduplex formationreaction. The appearance of high-molecular-mass DNA struc-tures after purification of the ds PCR products and afterheteroduplex formation has not previously been described. Theformation of Holliday junctions before CCM or EMC analysismay explain some of the background cleavage, as T4endonuclease VII is an enzyme capable of resolving four-wayDNA structures [27]. It has also been previously shown thatHolliday junctions may be cleaved chemically [24] andtherefore their formation may also cause the backgroundcleavage during CCM. Previously, the non-specificbackground cleavage in EMC analysis in all positions [8,10]was explained by the bulging or curved duplex structures [46].Our results indicate that it may also be caused by the formationof four-way DNA structures.

To optimize CCM and EMC methods, the fidelity of thelinear heteroduplex needs to be improved by decreasing theformation of four-stranded DNA. This may be achieved inthe following ways: using DNA concentrations lower than5 ng´mL21; analysing shorter DNA fragments overlapping theregion of interest; purifying heteroduplexes before mutationdetection analysis and preincubating DNA solution of lowconcentration to resolve any x structures formed during DNAprecipitation; creating C-G clamps by using primers withadditional G and C bases on the 5 0 ends to prevent melting ofdsDNA at the ends and therefore to decrease the probabilityof DNA±DNA interaction.

The value of the proposed approaches is confirmed by resultsalready obtained using different approaches to CCM optimiza-tion. The greater ability of longer DNA fragments to formtetramers and high-molecular-mass structures also correlateswith CCM analysis performed on the 1022-bp PCR product ofthe PS1 gene; a high background was consistently seen usingthese fragments. The use of smaller fragments overlapping theregion of interest improved CCM analysis results [47]. Ananalogous result was reported by Verpy et al. [48], who notedthat a 1325-bp product of the A region of the C1 INH gene wasunable to form perfect heteroduplexes and used several smallerfragments overlapping the A region of the C1 INH gene. Thesolid-phase CCM approach provides an increase in the relativeamount of double-stranded linear heteroduplexes directlybefore CCM analysis. Traditional CCM analysis of the PCRproduct of exon 5 of the p53 gene gave a dense smear at the topof a gel, which was greatly reduced in solid-phase CCManalysis [49].

Taken together, our results show an ability of homologousduplexes to interact, with the formation of Holliday junctions.The mechanism of DNA±DNA interaction proposed is inaccordance with modern theory of DNA duplex melting andfraying. This property of DNA may be important for under-standing DNA rearrangement mechanisms. As the list of uses ofPCR for scientific and diagnostic purposes grows, the possibleinteraction of homologous duplexes in PCR product solutionsmay also be of practical significance.

A C K N O W L E D G E M E N T S

We thank the following: P. M. Chumakov for kindly donating the set of

plasmids with p53 cDNA and for his very valuable theoretical advice; the

Neurogenetics and Neurotransgenesis Groups, Mayo Clinic, Jacksonville

for kindly giving us the plasmids with PS1 gene sequence; Dr Myers for


kindly donating the set of plasmids with mouse b-globin gene promoter

region. We are grateful to Z. Ronai, F. L. Kisseljov, A. V. Likhtenshtein,

A. Lyubimov and V. P. Shelepov for valuable discussions. We are also

grateful to Yu. Emelianov for technical assistance and V. Mehl for

manuscript preparation. This work was supported by RFFI grant No. 98-04-

48182 and 00-04-48579a, a CDA-NIS grant from NCI, USA to M.G.Y., an

ICREET Fellowship from UICC to M.G.Y., an Australian Postgraduate

Award to J.J.B. and a grant from the National Health and Medical Research

Council of Australia.

R E F E R E N C E S

1. Cotton, R.G.H. (1997). Mutation Detection. Oxford University Press,

Oxford, UK.

2. Ronai, Z. & Yakubovskaya, M.G. (1995) PCR in clinical diagnosis.

J. Clin. Lab. Anal. 9, 269±281.

3. Cotton, R.G.H. & Horaitis, Q. (2000) Quality control in the discovery,

reporting, and recording of genomic variation. Hum. Mutat. 15,

16±21.

4. Cotton, R.G.H., Rodrigues, N.R. & Campbell, R.D. (1988)

Reactivity of cytosine and thymine in single-base-pair

mismatches with hydroxylamine and osmium tetroxide and its

application to the study of mutations. Proc. Natl Acad. Sci. USA 85,

4397±4403.

5. Lambrinakos, A., Humphrey, K.E., Babon, J.J., Ellis, T.P. &

Cotton, R.G.H. (1999) Reactivity of potassium permanganate

and tetraethylammonium chloride with mismatched bases and a

simple mutation detection protocol. Nucleic Acids Res. 27,

1866±1873.

6. Cotton, R.G.H. & Campbell, R.D. (1989) Chemical reactivity of

matched cytosine and thymine bases near mismatched and

unmatched bases in a heteroduplex between DNA strands with

multiple differences. Nucleic Acids Res. 17, 4223±4233.

7. Lu, A.-L. & Hsu, I.-C. (1992) Detection of single DNA base

mutations with mismatch repair enzymes. Genomics 14, 249±253.

8. Youil, R., Kemper, B. & Cotton, R.G.H. (1995) Screening for

mutations by enzyme mismatch cleavage with T4 endonuclease VII.

Proc. Natl Acad. Sci. USA 92, 87±93.

9. Youil, R., Kemper, B. & Cotton, R.G.H. (1996) Detection of 81 of 81

known mouse beta-globin promoter mutations with T4 endo-

nuclease VII: the EMC method. Genomics 32, 431±437.

10. Mashal, R.D., Koontz, J. & Sklar, J. (1995) Detection of mutations by

cleavage of DNA heteroduplexes with bacteriophage resolvases.

Nat. Genet. 9, 177±182.

11. Gaillard, C. & Strauss, F. (1994) Association of poly (CA) poly (TG)

DNA fragments into four-stranded complexes bound by HMG1 and

2. Science 264, 433±436.

12. Belotserkovskii, B.P. & Johnston, B.H. (1996) Polypropylene tube

surfaces may induce denaturation and multimerization of DNA.

Science 271, 222±223.

13. Gaillard, C. & Strauss, F. (1996) Polypropylene tube surfaces may

induce denaturation and multimerization of DNA-Response. Science

271, 223.

14. Belotserkovskii, B.P. & Johnston, B.H. (1997) Denaturation and

association of DNA sequences by certain polypropylene surfaces.

Anal. Biochem. 251, 251±262.

15. Yakubovskaya, M.G., Neschastnova, A.A., Popenko, V.I., Lipatova,

Zh.V. & Belitsky, G.A. (1999) Holliday junctions are formed in

concentrated solutions of purified products of DNA amplification.

Biochemistry (Moscow) 64, 1310±1314.

16. Gaillard, C., Flavin, M., Woisard, A. & Strauss, F. (1999) Association

of double-stranded DNA fragments into multistranded DNA

structures. Biopolymers 50, 679±689.

17. Panyutin, I.G. & Hsieh, P. (1994) The kinetics of spontaneous DNA

branch migration. Proc. Natl Acad. Sci. USA 91, 2021±2025.

18. Holliday, R. (1964) A mechanism for gene conversion in fungi.

Genet. Res. Camb. 5, 282±287.

19. Potter, H. & Dressler, D. (1976) On the mechanism of genetic

recombination: electron microscopic observation of recombination

intermediates. Proc. Natl Acad. Sci. USA 73, 3000±3004.

20. Lopez, B., Rousset, S. & Coppey, J. (1987) Homologous

recombination intermediates between two duplex DNA catalysed

by human cell extracts. Nucleic Acids Res. 15, 5643±5655.

21. Wasserman, S.A. & Cozzarelli, N.R. (1986) Biochemical topology:

applications to DNA recombination and replication. Science 232,

951±960.

22. Whithouse, H.L.K. (1982). Genetic Recombination: Understanding

the Mechanism. Wiley, New York, USA.

23. Cox, M.M. & Lehman, I.R. (1987) Enzymes of general recombina-

tion. Annu. Rev. Biochem. 56, 229±254.

24. Lilley, D.M.J. & Clegg, R.M. (1993) The structure of the four-way

junction in DNA. Annu. Rev. Biophys. Biomol. Struct. 22, 299±328.

25. Lilley, D.M.J. (1997) All change at Holliday junction. Proc. Natl

Acad. Sci. USA 94, 9513±9515.

26. Duckett, D.R., Murchie, A.I., Diekmann, S., von Kitzing, E.,

Kemper, B. & Lilley, D.M. (1988) The structure of the Holliday

junction, and its resolution. Cell 55, 79±89.

27. Mizuuchi, K., Kemper, B., Hays, J. & Weisberg, R.A. (1982) T4

endonuclease VII cleaves Holliday structures. Cell 29, 357±365.

28. Buchman, V.L., Chumakov, P.M., Ninkina, N.N., Samarina, O.P. &

Georgiev, G.P. (1988) A variation in the structure of the protein-

coding region of the human p53 gene. Gene 30, 245±252.

29. Smith, M.J., Gardner, R.J., Knight, M.A., Forrest, S.M., Beyreuther,

K., Storey, E., McLean, C.A., Cotton, R.G.H., Cappal, R. &

Masters, C.L. (1999) Early-onset Alzheimer's disease caused by a

novel mutation at codon 219 of the presenilin-1 gene. Neuroreport

25, 503±507.

30. Sambrook, J., Fritsch, E.F. & Maniatis, T. (1989) Molecular Cloning:

a Laboratory Manual. 2nd edn. Cold Spring Harbor Laboratory

Press, Cold Spring Harbor, NY, USA.

31. Southern, E.M. (1975) Detection of specific sequences among DNA

fragments separated by gel electrophoresis. J. Mol. Biol. 96,

503±517.

32. Davis, R.W. & Hyman, R.W. (1971) A study in evolution: the DNA

base sequence homology between coliphages T7 and T3. J. Mol.

Biol. 62, 287±301.

33. Lu, M., Guo, Q., Marky, L.A., Seeman, N.C. & Kallenbach, N.R.

(1992) Thermodynamics of DNA branching. J. Mol. Biol. 223,

781±789.

34. Duguid, J.G. & Bloomfield, V.A. (1995) Aggregation of melted DNA

by divalent metal ion-mediated cross-linking. Biophys. J. 69,

2642±2648.

35. Henderson, E., Hardin, C.C., Wolk, S.K., Tinoco, I. & Blackburn,

E.H. (1987) Telomeric DNA oligonucleotides form novel intra-

molecular structures containing guanine-guanine base pairs. Cell 51,

899±908.

36. Hardin, C.C., Henderson, E., Watson, T. & Prosser, J.K. (1991)

Monovalent cation induced structural transitions in telomeric DNAs:

G-DNA folding intermediates. Biochemistry 30, 4460±4472.

37. Mueller, J.E., Newton, C.J., Jensch, F., Kemper, B., Cunningham,

R.P., Kallenbach, N.R. & Seeman, N.C. (1990) Resolution of

Holliday junction analogs by T4 endonuclease VII can be directed

by substrate structure. J. Biol. Chem. 265, 13918±13924.

38. Muck, S.M., Fee, R.S., Millar, D.P. & Chazin, W.J. (1997) Crossover

isomer bias is the primary sequence-dependent property of

immobilised Holliday junctions. Proc. Natl Acad. Sci. USA 94,

9080±9084.

39. Saenger, W. (1984). Principles of Nucleic Acid Structure. Springer-

Verlag, New York, USA.

40. Nonin, S., Leroy, J.L. & Gueron, M. (1995) Terminal base pairs of

oligodeoxynucleotides: imino proton exchange and fraying.

Biochemistry 34, 10652±10659.

41. Renalldo, R.P., Vologodskii, A.V., Neri, B.P. & Lyamichev, V.I. (2000)

The kinetics of oligonucleotide replacements. J. Mol. Biol. 297,

511±520.

42. Boan, F., Rodriguez, J.M. & Gomez-Marquez, J. (1998) A non-

hypervariable human minisatellite strongly stimulates in vitro


intramolecular homologous recombination. J. Mol. Biol. 278,

499±492.

42a. Eichman, B.F., Vargason, J.M., Mooers, B.H. & Ho, P.S. (2000) The

Holliday junction in an inverted repeat DNA sequence: sequence

effects on the structure of four-way junctions. Proc. Natl Acad. Sci.

USA 97, 3971±3976.

43. Benet, A. & Azorin, F. (1999) The formation of triple-stranded

DNA prevents spontaneous branch-migration. J. Mol. Biol. 294,

851±857.

44. Frank-Kamenetskii, M.D. & Mirkin, S.M. (1995) Triplex DNA

structures. Annu. Rev. Biochem. 64, 65±95.

45. Gierer, A. (1966) Model for DNA and protein interactions and

function of the operator. Nature (London) 212, 1480±1481.

46. Bhattacharyy, A., Murchie, A.I., von Kitzing, E., Diekmann, S.,

Kemper, B. & Lilley, D.M. (1991) Model for the interaction of

DNA junctions and resolving enzymes. J. Mol. Biol. 221,

1191±1207.

47. Smith, M.J., Humphrey, K.E., Cappai, R., Beyreuther, K.,

Masters, C.L. & Cotton, R.G. (2000) Correct heteroduplex

formation for mutation detection analysis. Mol. Diag. 5, 67±73.

48. Verpy, E., Biasotto, M., Brai, M., Misiano, G., Meo, T. & Tosi, M.

(1996) Exhaustive mutation scanning by fluorescence-assisted

mismatch analysis discloses new genotype±phenotype correlations

in angiodema. Am. J. Hum. Genet. 59, 308±319.

49. Hansen, L.L., Justesen, J. & Kruse, T.A. (1996) Sensitive and fast

mutation detection by solid phase chemical cleavage. Hum. Mutat.

7, 256±263.


Interaction of linear homologous DNA duplexes via Holliday junction formation

Documents