Specific Mismatch Recognition in Heteroduplex Intermediates by p53 Suggests a Role in Fidelity Control of Homologous Recombination

Abstract

We demonstrate that wild-type p53 inhibits homologous recombination. To analyze DNA substrate specificities in this process, we designed recombination experiments such that coinfection of simian virus 40 mutant pairs generated heteroduplexes with distinctly unpaired regions. DNA exchanges producing single C-T and A-G mismatches were inhibited four- to sixfold more effectively than DNA exchanges producing G-T and A-C single-base mispairings or unpaired regions of three base pairs comprising G-T/A-C mismatches. p53 bound specifically to three-stranded DNA substrates, mimicking early recombination intermediates. The K_D values for the interactions of p53 with three-stranded substrates displaying differently paired and unpaired regions reflected the mismatch base specificities observed in recombination assays in a qualitative and quantitative manner. On the basis of these results, we would like to advance the hypothesis that p53, like classical mismatch repair factors, checks the fidelity of homologous recombination processes by specific mismatch recognition.

p53 germ line mutations are associated with a deficit to maintain genomic stability along with an increase of spontaneous gene amplification rates (17, 52, 93), thereby accelerating the multistep process of tumor progression (81). This phenotype has been explained by the loss of p53 cell cycle checkpoint control (38, 39, 46). DNA damage (38, 39, 60) and suboptimal growth situations, such as an increase of oxygen radicals (28) or ribonucleotide depletion (50), are signals for p53-mediated accumulation and functional activation (54, 68). Depending on the cell type, p53 induces cell cycle arrest or apoptosis predominantly via transcriptional transactivation of genes coding for the cyclin-dependent kinase inhibitor p21/WAF1/CIP1/SDI1 (21, 23) or the apoptotic factor Bax (56). As a consequence, cells are unable to replicate their DNA under conditions which may lead or may have led to chromosome breaks (3), thereby preventing the manifestation and aggravation of genomic lesions in S phase.

Strikingly, the same molecular signal triggering the DNA damage response by p53, namely, DNA strand breaks (60), also initiates V(D)J recombination (79), meiotic recombination (27), recombination repair (75), and gene amplification (19) events. There is evidence for an at least indirect involvement of p53 in V(D)J recombination, as γ irradiation can rescue rearrangement at multiple T-cell receptor loci by a p53-dependent bypass mechanism in scid mice (2, 12). A role for p53 in meiotic recombination has been postulated from the observation that p53 mRNA expression in testes of mice is high and specific for spermatocytes in zygotene to pachytene, the meiotic stages at which homologous chromosomes synapse for genetic exchange (65, 69). Intriguingly, the mitotic checkpoint factor Atm, the product of the gene mutated in patients with ataxia telangiectasia (66), is also found in spermatocytes of meiosis I. Atm belongs to the family of phosphatidylinositol 3-kinase-like protein kinases which, like DNA-dependent kinase DNA-PK, another phosphatidylinositol 3-kinase family member involved in V(D)J recombination and so-called end-joining pathways of double-strand break (DSB) repair (for reviews, see references 36 and 85), are good candidates for signal-amplifying molecules after sensing DNA aberrations (18). Mutations in ATM lead to a delay in the response of p53 toward ionizing radiation, indicating upstream functions within the signaling response (39, 53).

Further support for the idea that p53 is a cell cycle-regulatory factor also directly linked to repair and/or recombination processes comes from a number of reports on physical interactions with proteins involved in DNA-modifying pathways: replication factor A, a single-stranded DNA (ssDNA) binding protein participating in DNA replication, DNA damage recognition, recombination, and nucleotide excision repair (reference 70 and references therein); XPB, XPD, and p62, two helicases and one subunit of unknown function of the dual transcription/excision repair complex TFIIH (84, 90); and the human RecA homolog Rad51 (73), which performs homologous DNA pairing and strand exchange reactions with a polarity opposite that of RecA (10, 26, 74). Intriguingly, homozygous Rad51 knockouts show early embryonic death, which can be alleviated by a mutation in p53 (49, 77). In addition, p53 itself performs biochemical activities, such as DNA reannealing and strand transfer on short oligonucleotides (4, 15, 61), and 3′-5′ exonuclease activity (59), which suggests active participation in repair processes possibly comprising homologous recombination. The C-terminal 30 amino acids of p53 seem to function like a molecular switch regulating the functions of the p53 core domain, as could be demonstrated by C-terminal phosphorylation, truncation of the protein, or antibody activation (34, 35, 37, 59). The same region on p53 also binds ssDNA sequence nonspecifically (5), recognizes insertion/deletion-type DNA mismatches, and confers DNA damage sensor functions by stimulating sequence-specific DNA binding of the central part (37, 47, 63). Strong indications for a direct involvement of p53 in genetic exchange come from two recent publications showing DNA conformation-specific recognition of cruciform (42) and Holliday junction (48) DNAs in vitro.

Favoring the idea of a direct regulatory role of p53 in DNA exchange processes, we have developed a model system which allows us to quantify homologous recombination rates between simian virus 40 (SV40) chromosomes in monkey cells (89) and avoids unwanted effects originating from p53 responses to nakedly transfected DNA (68). This assay is based on measuring the recombination rates between two types of SV40 whose genomes were mutated in such a way that upon double infection of monkey cells, virus particles could be released only after interchromosomal exchange of genetic material. By use of this system, we were able to demonstrate suppression of homologous recombination events by wild-type p53, which could be alleviated by complexing p53 with SV40 T antigen (T-Ag). Our interpretation of a direct involvement of p53 in the control of homologous recombination has been supported by recent reports describing an increase in spontaneous intrachromosomal homologous recombination upon functional inactivation of wild-type p53 (11, 55).

In this study, we have applied our SV40-based test system to the analysis of specific heteroduplex DNA substrates, with respect to the inhibition of homologous recombination events by p53. We demonstrate a preferred inhibition for certain base-base mismatch types in heteroduplexes produced by DNA exchange in living cells, which correlates with the binding of artificial recombination intermediates by p53 in vitro.

MATERIALS AND METHODS

Mammalian cells and transgenic clones.

Kidney cells from African green monkey (Cercopithecus aethiops), CV1, TC7, and rhesus monkey (Macaca mulatta) kidney cells, from primary isolates (PRK), and from the established cell line LLC-MK₂ (33, 82) were grown in Dulbecco modified Eagle medium (DMEM) supplemented with 10% fetal calf serum (FCS) at 37°C. LLC-MK₂(p53her) cells were established after calcium phosphate-mediated cotransfection of LLC-MK₂ cells with pSV53her (64) and pM5neo; LLC-MK₂(neo) cells were established after transfection with pM5neo only. Clones were raised in DMEM supplemented with 0.8 mg of G418 (Gibco/BRL) and 10% FCS, stripped by stirring 1 liter with 10 g of charcoal (Norit A; Merck) and 1 g of Dextran 40 (Merck) for 30 min, centrifuged at 13,000 × g, and passed through 0.2-μm-pore-size filters. For analysis and assays, the cells were kept in phenol red-free DMEM with charcoal-stripped FCS. β-Estradiol (Sigma) was added to result in the final concentrations indicated. Isolated clones were tested for p53her (see Results for description) expression in total homogenates by Western transfer and immunodetection.

SV40 and recombination assays.

Genomic SV40 (strain 776 derivative) and SV40-tsB11 (76) DNAs were isolated, linearized by KpnI, and cloned into the KpnI site of pBluescript M13+ (Stratagene), using competent Escherichia coli XL1Blue (Stratagene). The complete coding region of VP1 in pBS-SV40-tsB11 was sequenced by using a T7 polymerase kit from Pharmacia. The 273-bp ApaI-BamHI DNA fragment encompassing the tsVP1(K290T)/A2367C mutation was transferred into the corresponding sites of pUC-SV40 (89), thereby creating pUC-SV40-tsVP1(K290T). pBS-SV40-tsVP1(286S)-C2354T was engineered by in vitro mutagenesis on ssDNA of pBS-SV40 prepared after M13KO7 helper phage (Pharmacia) infection of the corresponding XL1Blue bacteria. For that purpose, we used the Sculptor in vitro mutagenesis system (Amersham Buchler) according to the manufacturer’s instructions. The mutagenizing oligonucleotide was 5′-GCAGTGGAAGGGACTTTCCAGATATTTTAAAATTACC-3′ (mutating base is underlined). The absence of extra mutations was excluded by sequencing and ApaI-BamHI recloning of the mutated fragment into pUC-SV40. DNA-modifying enzymes were purchased from Boehringer Mannheim, New England Biolabs, Fermentas, or Pharmacia. Virus generation from recombinant viral DNAs, virus infection, and determination of virus yields (multiplicity of infection) via T-Ag immunofluorescence were performed exactly as described by Wiesmüller et al. (89). Recombination assays were performed and evaluated accordingly, except that β-estradiol-containing medium was added to LLC-MK₂(p53her) and LLC-MK₂(neo) cells immediately after virus infection.

Extraction of mammalian cells, immunoprecipitation, and Western blot analysis.

After washing cells three times with phosphate-buffered saline (PBS), we obtained total cellular homogenates at 0°C by scraping the cells from the culture dish with a rubber policeman in 100 μl of 3× sodium dodecyl sulfate (SDS) sample buffer (65 mM Tris-HCl [pH 6.8], 10% glycerol, 2.3% SDS, 5% β-mercaptoethanol, bromophenol blue) per 10⁶ cells. Protocols for cell extractions, immunoprecipitations, and Western blot analyses were performed as described by Wiesmüller et al. (89). For anti-p53 immunoprecipitations from cell extracts of different LLC-MK₂(p53her) and LLC-MK₂(neo) clones untreated or after treatment with 5 μM β-estradiol for 24 h, we used the wild-type p53 conformation-specific monoclonal antibody Pab1620 (9). For the corresponding virus infections, we used SV40-dl1066, which encodes T-Ag lacking the C-terminal host range domain of 38 amino acids, in order to avoid signal interference in immunoblots, as T-Ag has an apparent molecular mass of 90 kDa, identical to that of p53her. At 48 h after infection and culturing in the absence or presence of 5 μM β-estradiol, T-Ag–p53 complexes were isolated by precipitation with Pab419, which is directed against the N terminus of T-Ag (88). Proteins separated on an SDS–10% polyacrylamide gel and transferred to Hybond-C Super membranes (Amersham) were immunodetected by using sheep polyclonal anti-p53 serum (Boehringer Mannheim) and the affinity-purified and peroxidase-conjugated secondary serum which is goat anti-sheep immunoglobulin IgG (Sigma). Visualization of the immunocomplexed p53 bands was achieved by chemiluminescence enhancement according to the Amersham protocol.

Assay for transcriptional transactivation.

LLC-MK₂(p53her) and LLC-MK₂(neo) cells were transiently lipofected with the p53 reporter plasmid pG13-CAT (40) by use of Lipofectamine (Gibco/BRL). After a recovery phase of 24 h, the cells from one culture dish each were split, subjected to a recovery phase of 8 h, and further cultivated for 16 h in the presence or absence of 2 μM β-estradiol. In parallel, LLC-MK₂ cells stably expressing the Gal4-estrogen receptor fusion protein GalER-VP16 were lipofected with the Gal4 reporter pCAT-4 and estradiol induced correspondingly, resulting in comparable transactivation activities (16). As a positive control, LLC-MK₂ cells were lipofected with pSV2CAT, which directs constitutive expression of the chloramphenicol acetyltransferase (CAT) enzyme from the early SV40 promoter. During the preparation of cell extracts 48 h after lipofection and for the quantitative determination of CAT in these extracts, we followed the instructions accompanying the CAT enzyme-linked immunosorbent assay kit from Boehringer Mannheim. To calculate the specific transcriptional transactivation activities in the extracts, we measured the protein contents by use of the bicinchoninic acid protein assay reagent (Pierce GmbH).

Baculoviral expression and preparation of p53 from insect cells.

High Five insect cells (Invitrogen) were infected either with recombinant baculovirus directing the expression of N-terminally histidine-tagged murine wild-type p53, of histidine-tagged p53Gly168,Ile234 mutant protein from MethA cells (20), or of human wild-type p53. Cells harvested 48 h after infection were washed four times with PBS at 4°C and incubated in ice-cold buffer A (10 mM HEPES [pH 7.4], 1.5 mM MgCl₂, 5 mM KCl, 1 mM dithiothreitol [DTT], protease inhibitors [125 μg of Pefabloc, 5 μg of pepstatin, 5 μg of leupeptin, and 5 μg of aprotinin per ml) for 60 min. Swollen cells were Dounce homogenized and left on ice for a further 45 min. Intact nuclei were harvested by centrifugation at 5,000 × g, resuspended in buffer B (10 mM HEPES [pH 9.0], 1.5 mM MgCl₂, 5 mM KCl, 10 mM DTT, protease inhibitors), and extracted for 45 min. After repeated centrifugation, the pellet was extracted with buffer B containing 200 mM KCl. This fraction contained the major portion of p53, as verified by immunoblot analysis. After SDS-polyacrylamide gel electrophoresis (PAGE) and Coomassie blue staining, the 53-kDa band corresponding to p53 appeared as the dominant band, indicating high enrichment of p53 in the preparation. p53-containing fractions were aliquoted and stored frozen at −70°C. The quality of each protein preparation was checked by electrophoretic mobility shift assays (EMSAs) with double-stranded oligonucleotides mimicking the wild-type p53-specific ribosomal gene cluster (RGC) site (40) and with DNA fragments corresponding to the mutant p53-specific matrix attachment region/scaffold attachment region (MAR/SAR) elements (58).

Wild-type p53 protein from insect cells was independently purified to homogeneity via affinity binding of the N-terminal tag consisting of six histidines to Talon metal resin from Clontech (59). The second method applied to obtain highly purified wild-type proteins, Pab421 immunoaffinity purification, differed from the published procedure by substituting 3.0 M MgCl₂ elution for alkaline elution. Subsequently, p53-containing fractions were dialyzed against 20 mM Tris (pH 7.5)–200 mM NaCl–2 mM DTT. The homogeneous p53 protein preparations (≥90% pure) were stored at 4°C and applied to band shift analysis using three-stranded substrates within 2 days. Protease inhibitors were purchased from Bayer, Biomol, or Serva.

Preparation of three-stranded intermediates.

Artificial three-stranded DNA substrate representing the perfectly matching recombination intermediate at the SV40-tsVP1(290T) locus was prepared by annealing top (5′-ACGCTGCCGAA TGGATCCGGTTATCACCGCTTTCTAAGGGTAATTTTAAAATATCTGG GAAGTCCC-3′), central (5′-GACTTCCCAGATATTTTAAAATTACCCTT AGAAAGCGGTCTGTGAAAAACCCCTACCCAATTTCCTT-3′), and bottom (5′-GGAAATTGGGTAGGGGTTTTTCACAGACCGCTTTCTAAGCT GTCTAGAGGATCCGACTATCGA-3′) oligonucleotides simultaneously. For the corresponding C-T mismatch substrate, central oligonucleotide 5′-GACTT CCCAGATAT TT TACAAT TACCC T TAGAAAGCGGTC TGTGAAAAACC CCTACCCAATTTCCTT-3′ (mispaired base underlined) was used instead. The matching intermediate at the SV40-tsVP1(286S) locus was mimicked by junction DNAs consisting of top (5′-GACGCTGCCGAATGGATCCGGTTAAAGGGT AATTTTAAAATATCTGGGAAGTCCCTTCCACTGCTG-3′), central (5′- GCAGTGGAAGGGACTTCCCAGATATTTTAAAATTACCCTTAGAAA GCGGTCTGTGAAAAACCCCTA-3′), and bottom (5′-GGGGTTTTTCACA GACCGCTTTCTAAGGGTAATTTTAAAGCTGTCTAGAGGATCCGAC TATCGA-3′) oligonucleotides. At the same locus, mismatch substrates were produced by incorporating top oligonucleotide 5′-GACGCTGCCGAATGGAT CCGGTTAAAGGGTAATTTTAAAATATCTGGAAAGTCCCTTCCACTG CTG-3′ for A-C mispairing, top oligonucleotide 5′-GACGCTGCCGAATGGATCCGGTTAAAGGGTAATTTTAAAATATCTGGTAAGTCCCTTCCACTGCTG-3′ for C-T mispairing, central oligonucleotide 5′-GCAGTGGAAGGGAC TTTCCAGATATTTTAAAATTACCCTTAGAAAGCGGTCTGTGAAAAAC CCCTA-3′ for G-T mispairing, central oligonucleotide 5′-GCAGTGGAAGGG ACTTACCAGATATTTTAAAATTACCCTTAGAAAGCGGTCTGTGAAA AACCCCTA-3′ for A-G mispairing, and central oligonucleotide 5′-GCAGTGG AAGGGACTTTCTAGATATTTTAAAATTACCCTTAGAAAGCGGTCTG TGAAAAACCCCTA-3′ for G-T/G-T mispairings. In each case, top oligonucleotides were radioactively labeled at the 5′ end prior to annealing by use of [γ-³²P]ATP and T4 polynucleotide kinase. Labeled and unlabeled competitor junction DNAs were separated from flayed duplexes and ssDNAs by native electrophoresis in a TBE (89 mM Tris-HCl [pH 8.3], 89 mM boric acid, 2.5 mM EDTA)-buffered 6% polyacrylamide gel, their positions were identified by autoradiography, and they were eluted from crushed gel pieces under agitation at 4°C for 16 h. DNA concentrations were measured by using DNA Dipsticks (Invitrogen), and the resulting values were controlled by scintillation counting of the corresponding radioactivities.

EMSA and Scatchard analysis.

³²P-labeled junction, flayed duplex, or ssDNAs were mixed with p53 protein at the concentrations indicated for each figure in 25 mM Tris-HCl (pH 8.0)–5 mM EDTA–1 mM DTT–6% glycerol (shift buffer) in a total volume of 20 μl. In control reactions without p53, the protein storage buffer only was added. For supershifts, 1 μl of a protein fraction containing 100 ng of p53 from insect cells was mixed with shift buffer, then 100 ng of the corresponding affinity-purified monoclonal antibody in a volume of 1 μl of PBS was added, and the mixture was preincubated for 45 min before the addition of DNA. Controls without antibody contained 1 μl of PBS instead. After incubation for 25 min (1 to 20 min for on-rate studies) on ice, the protein-DNA mixtures were subjected to PAGE on a native 4% polyacrylamide gel in 6.7 mM Tris-HCl (pH 8.0)–3.3 mM sodium acetate–2 mM EDTA at room temperature. Dried gels were autoradiographed, and band intensities were quantified by PhosphorImager analysis. Individual backgrounds were subtracted for each lane individually. In each gel, the intensities of the bands at the position of three-stranded DNA from control reactions with input DNAs of known concentrations but lacking protein were used to calculate the DNA concentrations of bound and free DNAs. The free substrate concentration was estimated as the product of the fraction of ³²P-labeled DNA in the unshifted position and the total concentration of input DNA. The total bound substrate concentration was estimated as the product of the fraction of ³²P-labeled DNA in the shifted position and the total concentration of input DNA. The apparent equilibrium dissociation constant K_D and the corresponding standard error were calculated from the Scatchard plot by computational analysis via the program Grafit. Three-stranded substrate concentrations in the final Scatchard analysis were 44 to 612 pM for match, 15 to 306 pM for C-T and A-G, and 230 to 2,295 pM for G-T, A-C, and G-T/G-T. Off rates of labeled three-stranded DNAs bound to p53 were determined after addition of either 2 × 10²-fold excess of nonlabeled specific DNA or 3 × 10⁵-fold excess of competitor ssDNA.

RESULTS

Reduction of recombination rates in monkey cells by ectopically expressed wild-type p53.

For the analysis of p53 functions in recombination between SV40 chromosomes, we decided to establish an SV40-infectible monkey cell line which displays wild-type p53 activities conditionally. As a consequence, cultivation should be possible under permissive conditions despite p53’s antiproliferative effect. To facilitate the interpretation of our data, we chose a monkey cell line without endogenous wild-type p53, the rhesus monkey cell line LLC-MK₂ (33). This cell line carries mutated p53 lacking amino acids 237 to 239 (p53Δ237-239) and as a consequence displays all of the characteristics of a transforming p53 mutant (82). Prior experiments demonstrated that p53Δ237-239 had lost recombination suppressor activities completely (89).

LLC-MK₂ cells were transfected with the construct pSV53her (64) for the stable expression of p53her (human wild-type p53 C-terminally fused to the human estrogen receptor hormone binding domain), which allows activation of wild-type p53 functions after 17β-estradiol application. Importantly, within this chimera, the C-terminal oligomerization domain of p53 was not expected to be accessible to other proteins before estradiol addition, as hormone binding by the human estrogen receptor fusion liberates this peptide and the neighboring C-terminal regions of p53 from a protein complex containing heat shock proteins (13, 45, 67). Accordingly, mutant p53Δ237-239 should exist predominantly within homo-oligomers in the absence of β-estradiol. Therefore, dominant negative effects of p53Δ237-239 via hetero-oligomerization with p53her after estradiol induction can largely be excluded. Another advantage of ectopic p53her expression in view of the SV40-based recombination assay was the fact that significant amounts of this wild-type p53 fusion protein would be active before T-Ag synthesis and complex formation of p53her after SV40 infection (89).

When we examined cells from different LLC-MK₂(p53her) clones for de novo synthesis and steady-state levels of p53her (data not shown), the hybrid protein with an apparent molecular mass of 90 kDa in SDS gels was found to be maximally expressed in clones 17 and 29; therefore, we selected these clones for more detailed analysis and most of the subsequent recombination studies. According to computer-assisted quantitative evaluation of autoradiographs, the level of p53her protein was in the physiological range of endogenous wild-type p53 from primary rhesus monkey or TC7 cells and threefold below the level of endogenous p53Δ237-239 in LLC-MK₂ cells, both of which did not change after application of 10 μM β-estradiol for 4 h (Fig. 1A). As depicted in Fig. 1B, p53her was immunoprecipitated by the wild-type p53 conformation-specific monoclonal antibody Pab1620 (9). Immunoprecipitations with Pab419 (88) directed against the T-Ag gave evidence for complex formation of p53her and SV40 T-Ag 48 h after SV40 infection (Fig. 1C). Treatment of cells with 5 μM β-estradiol raised the fraction of p53her protein accessible to Pab1620 and T-Ag three- and twofold, respectively. However, Fig. 1B and C also demonstrate reactivity to p53her before estradiol addition, reflecting the basal level of active p53 protein due to residual estrogens in phenol red-free and hormone-stripped culture medium. The existence of functional wild-type p53 before hormone induction can also be deduced from the prolonged generation time of 27 h for LLC-MK₂(p53her) cells versus 21 h for LLC-MK₂(neo) cells. Transcriptional transactivation via the p53-responsive RGC element (40) has been demonstrated for p53her by others (64). Here, the production of the CAT reporter protein after transient lipofection of LLC-MK₂(p53her) cells with pG13-CAT containing 13 RGC elements was stimulated sixfold by incubation with 2 μM estradiol for 16 h (data not shown).

FIG. 1.

Western analysis of total and functionally inducible p53her. (A) Green monkey (TC7), primary rhesus monkey (PRK), neomycin-resistant control rhesus monkey [LLC-MK₂(neo)] and p53her-expressing rhesus monkey kidney cells [LLC-MK₂(p53her)] were cultivated in the absence (−) or presence (+) of 10 μM β-estadiol for 4 h. Total homogenates of 10⁵ cells each were subjected to SDS-PAGE and Western transfer, and p53 protein was detected by sheep anti-p53 serum, anti-sheep immunoglobulin G-peroxidase conjugate, and chemiluminescent peroxidase substrate. After incubation of LLC-MK₂(p53her)-17 cells for 24 h (B) or for 48 h after SV40 infection (C) with or without 5 μM β-estradiol, p53her was immunoprecipitated with Pab1620 or in complex with T-Ag via Pab419, respectively. Precipitates from 5 × 10⁵ cells each were treated for immunoblot analysis as described above.

Having established that p53her displays bona fide wild-type p53 features in LLC-MK₂(p53her) cells, we applied our SV40-based recombination assay system to LLC-MK₂(p53her)-17 cells and LLC-MK₂(neo) control cells (89). This system relies on differently mutated SV40 variants, which display a temperature-sensitive (ts) phenotype, such that production of virus particles is still possible at 32°C. After coinfection of monkey cells at the restrictive temperature of 39°C with two distinct SV40-tsVP1 mutants, virus particles can be generated only after exchange of genetic material and reconstitution of the wild-type genome. The results from five independent experiments using SV40-tsVP1(196Y) and SV40-tsVP1(286S) for coinfections showed that recombination frequencies in LLC-MK₂(p53her)-17 cells before ([3.5 ± 1.0] × 10⁻⁵) and after ([2.1 ± 0.9] × 10⁻⁵) β-estradiol addition did not change significantly, as was expected from the high basal level of wild-type p53 functions. Therefore, in the following experiments we decided to compare only recombination rates after wild-type p53 activation in LLC-MK₂(p53her) cells in the presence of β-estradiol and after the same treatment of control cells. LLC-MK₂(neo) cells allowed genetic exchange at a rate of (7.5 ± 2.1) × 10⁻⁵ in the presence of estradiol (five different experiments). To exclude false interpretations originating from clonal differences, we performed recombination measurements for LLC-MK₂(p53her)-17, LLC-MK₂(p53her)-29, and LLC-MK₂(p53her)-25 cells and cells from two independent LLC-MK₂(neo) clones and obtained comparable values (2 × 10⁻⁵, 4 × 10⁻⁵, 10⁻⁵, 8 × 10⁻⁵, and 8 × 10⁻⁵, respectively). On average, we saw a fourfold inhibition of homologous recombination events between SV40 chromosomes after expression and functional induction of wild-type p53 in monkey cells carrying a mutationally inactivated endogenous p53.

Recombination-regulating activities of p53 are influenced by certain mismatch types in heteroduplexes.

From closer examination of the exchange process provoked by use of the virus combination SV40-tsVP1(196Y) and SV40-tsVP1(286S), it becomes clear that heteroduplexes formed between the two genomes comprise two unpaired regions, which are separated by 269 perfectly complementary base pairs (89). Within the two possible heteroduplexes, a single-base mismatch of the A-C or G-T type would be predicted at the VP1(196Y) locus, and two A-C or two G-T mismatches interrupted by one matching base pair would be predicted at the VP1(286S) locus (Fig. 2). Thus, both mutations C2084T [tsVP1(196Y)] and C2354T,C2356T [tsVP1(286S)] cause mismatches of the same base-base combination in recombination intermediates. To study the possible influence of a different mismatch type, we introduced the mutation A2367C in the immediate neighborhood of the VP1(286S) locus within the coding region for the late viral protein VP1 of SV40. This gives rise to A-G/C-T mismatches in heteroduplexes with the corresponding wild-type sequence in SV40-tsVP1(196Y). Like tsVP1(286S), this mutation, tsVP1(290T), causes a ts phenotype with respect to virus production and again does not allow cross-complementation with SV40-tsVP1(196Y) in trans (89), enabling us to utilize SV40-tsVP1(290T) for our infection-based recombination assay. As shown in Fig. 2, DNA exchange between SV40-tsVP1(196Y) and SV40-tsVP1(290T) genomes in LLC-MK₂(neo) were 36-fold as frequent as for the SV40-tsVP1(196Y)/SV40-tsVP1(286S) pair. In LLC-MK₂(p53her)-17, the cells we chose for measuring the recombination rates with different virus combinations, this high rate was reduced 21-fold after activation of p53her by 2 to 5 μM β-estradiol. In comparison to the fourfold suppression with SV40-tsVP1(196Y)/SV40-tsVP1(286S), this finding indicates that p53 inhibits recombination in a heteroduplex substrate-specific manner and strongly suggests a critical role of A-G/C-T mismatches in this process. The double mutation C2354T,C2356T in SV40-tsVP1(286S) will result in an unpaired region of three base pairs in recombination intermediates with SV40-tsVP1(196Y) DNA, whereas mutation A2367C in SV40-tsVP1(290T) will generate a single-base mismatch. To distinguish between possible influences of the base-base mismatch type and the size of the unpaired region, an SV40 mutant generating G-T/A-C single mismatches had to be analyzed. The mutation C2354T responsible for the codon alteration P286S lacking the silent C2356T exchange served this purpose and preserved the possibility of ts virus production. In recombination assays, this SV40-tsVP1(286S)-C2354T variant, showed similar exchange frequencies as SV40-tsVP1(286S) in both estradiol-treated LLC-MK₂(neo) and LLC-MK₂(p53her) cells, resulting in a fivefold inhibition by p53her. The fact that the two tsVP1(286S) variants caused similar degrees of inhibition of recombination rates in estradiol-induced p53her cells excluded the possibility that the length of the unpaired region, namely, 1 bp in heteroduplexes with SV40-tsVP1(290T)-C2354T and 3 bp with SV40-tsVP1(286S) DNAs, would have a major effect on p53 recombination-regulating activities.

FIG. 2.

Testing p53-regulated recombination for a dependency on distinct heteroduplex types. Recombination assays using SV40-tsVP1(196Y) in different SV40 combinations were performed to provoke distinct mismatches in the two possible heteroduplexes upon exchange of either strand of the viral genome. Italic letters indicate mutant information; capital letters indicate mutated nucleotides which cause the ts phenotype of the virus. Recombination rates in LLC-MK₂(neo) and LLC-MK₂(p53her)-17 cells were measured in the presence of 2 to 5 μM β-estradiol with SV40-tsVP1(286S) in five independent experiments, with SV40-tsVP1(286S)-C2354 in four assays, and with SV40-tsVP1(290T) in three assays. Standard errors were calculated for the mean values as indicated.

tsB11 virus (76), which had originally been isolated after nitrosoguanidine mutagenesis due to the ts phenotype caused by the same A2367C nucleotide exchange artificially created in SV40-tsVP1(290T), undergoes genomic exchange at a rate of (3.8 ± 1.7) × 10⁻⁵ (n = 3) in LLC-MK₂(neo) cells. This 71-fold decrease compared to the recombinant SV40-tsVP1(290T) can easily be explained by the existence of 0.5% overall sequence divergence in tsB11 versus wild-type viral genomes, as we extrapolated from the sequence analysis of the VP1 coding region. Possibly even more detrimental to the success of DNA exchange was the fact that the silent mutation T2239G in tsB11 cuts the homologous region between the indicator mutations at positions 2084 and 2367 down to segments below the minimal homologous region described for mammalian cells (83). The maintenance of low recombination rates in the absence of functional wild-type p53 must be attributed to p53-independent factors engaged in the avoidance or dissolution of sequence divergence in recombination intermediates. Nevertheless, p53her inhibited recombination activities further by at least a factor of 4, to <8 × 10⁻⁶, where we reach the detection limit for LLC-MK₂(p53her) cells.

Wild-type p53 specifically recognizes three-stranded DNA junctions.

Our observation that wild-type p53 displays mismatch substrate specificities in the recombination control of living cells prompted us to test the possibility of direct mismatch recognition within recombination intermediates. p53 was reported to recognize extrahelical loop mismatches on double-stranded DNAs in vitro (47). For our purposes, we produced radioactively labeled three-stranded recombination intermediates in vitro, using 63- to 67-mer oligonucleotides, according to a method which has been successfully applied to the functional analysis of well-known recombination factors (87). The sequence environment was chosen such that the nucleotide sequence of the central oligonucleotide represented exactly the wild-type SV40 genome 18 bp upstream to 48 bp downstream of the VP1(290T) mutation at position 2367. The top oligonucleotide was complementary only to the first 39 nucleotides at its 3′ end; the bottom oligonucleotide was complementary only to the last 38 nucleotides at its 5′ end, which produces a flexible junction in the middle by shared homologies of 12 nucleotides (Fig. 3). Aside from this three-stranded match, a three-stranded C-T mismatch substrate was produced by use of a central oligonucleotide altered only at the position corresponding to bp 2367 in SV40. This design for the analysis of mismatch recognition during or shortly after heteroduplex formation took into consideration the polarity of Rad51-mediated strand transfer initiating at the 5′ end of the complementary strand in the duplex (10, 74).

FIG. 3.

Specificity of binding to three-stranded junction DNAs by wild-type p53 (wtp53). Radioactively labeled top oligonucleotide (1s), flayed duplexes (2s C-T), and junction DNAs (3s and 3s C-T) were prepared, purified, and quantified as described in Materials and Methods. C-T mismatches in 2s C-T and 3s C-T substrates are positioned at the tsVP1(K290T)/A2367C corresponding locus and surrounded by the corresponding SV40 sequences. Substrates at 200 pM each and protein preparations enriched for wild-type or MethA mutant p53 protein (25 nM p53 tetramers) were mixed, incubated for 25 min on ice, and electrophoresed on a 4% polyacrylamide gel. The positions in the autoradiograph of substrate bands and of the major band shift containing p53 tetramers are indicated by arrows and schematic illustrations.

When these three-stranded substrates were incubated with protein fractions highly enriched for murine wild-type p53 overproduced in insect cells and the mixture was subjected to PAGE, the band containing the radioactively labeled substrate was completely shifted to a discrete upper position, indicative of the DNA substrate being tightly complexed with p53 tetramers (48, 71, 72). In addition, a minor band was seen with even more slowly migrating complexes, probably containing octameric forms of p53 (Fig. 3), as identified for sequence-specific binding (71). This picture did not emerge when wild-type p53 fractions were mixed with the corresponding two-stranded C-T mismatch substrate with flayed ends or the labeled top oligonucleotide at identical concentrations. Equal amounts of p53 MethA mutant protein (20) prepared in parallel did not retard any of the substrates to a discrete position in the gel. The protein interacted with DNA heterogeneously, as reflected by a smear along all four lanes. We noted the appearance of a new species of ssDNA migrating at the same position as double-stranded DNA after incubation with wild-type p53 fractions and much more pronounced with MethA p53. This phenomenon might be related to DNA-bending activities by wild-type p53 (8), which were also postulated for mutant p53 on A/T-rich DNA sequences (58), and will be discussed separately. Proof for the fully active state of each p53 protein was provided by testing the same protein fractions for sequence-specific DNA binding in the case of wild-type p53 (reviewed in reference 80) and MAR/SAR element binding in the case of MethA mutant protein (58, 86) (data not shown).

To prove the active participation of p53 in binding the three-stranded match and C-T mismatch substrates, we performed supershift analyses. Preincubation of protein fractions containing 0.5 pmol of wild-type p53 tetramers with 0.6 pmol of Pab246, specific for p53 in the wild-type conformation (92), resulted in the retardation of DNA-protein bands (Fig. 4A). Pab240, specific for p53 in the mutant conformation (24), had no effect. Consistent with an exclusive role of p53 in binding to the artificial recombination intermediates, His-tagged wild-type proteins purified from insect cells with the aid of Talon affinity chromatography or via Pab421 immunoaffinity to ≥90% homogeneity were each found to be fully active in substrate targeting (Fig. 4B). DNA binding was performed by p53 independently of the protein preparation, but differences between individual preparations were observed with respect to the multimerization state of p53 during binding. However, as can be seen in a protein dilution series (Fig. 4C), higher oligomeric forms of p53 in complex with three-stranded substrate can be obtained simply by increasing the protein concentrations. Comparable binding of three-stranded DNAs was performed by p53 from human origin (data not shown). In summary, our data allow us to draw the conclusion that wild-type p53 or a protein complex containing wild-type p53, but not the conformation-mutant MethA p53, recognizes imitations of early recombination intermediates strongly, specifically, and in distinct oligomeric complexes.

FIG. 4.

Analysis of wild-type p53 complexes in band shifts with three-stranded DNAs. Three-stranded substrates, as defined in the legend to Fig. 3, were applied to supershift analyses with monoclonal antibodies Pab246, specific for wild-type p53 (wtp53), and Pab240, specific for mutant p53 (A). Thin black arrows indicate the band shifts of p53-DNA complexes; thick shaded arrows indicate the supershifts. Homogeneous preparations of wild-type p53 after Talon or immunoaffinity purification were compared with p53-enriched protein preparations (standard) in EMSAs (B). Wild-type p53 was diluted in steps (50, 25, 12, and 6 nM calculated for the tetramer) in three-stranded DNA binding assays to examine the classes of multimeric complexes formed (C). If not indicated otherwise, p53 protein concentrations were 25 nM in the reaction mixtures with 200 pM junction DNAs in panel A, 100 pM in panel C, and 100 pM match substrate in panel B.

Different affinities of p53 to three-stranded intermediates are related to the type of integrated mismatch.

For a more detailed analysis of the binding characteristics, we measured affinities to several three-stranded DNA species differing only in their base-base mismatch type within the same sequence environment. To imitate the situation encountered in our preceding recombination assays as closely as possible, we used three-stranded junctions representing intermediates at the VP1(286S) locus, at which we had provoked mispairings in heteroduplexes by coinfection with different mutant virus pairs. So far, our in vitro experiments had demonstrated complete titration of three-stranded match and mismatch DNA inputs at concentrations of 200 pM by excess wild-type p53. To obtain equilibrium binding data for the determination of dissociation constants, DNA ligand concentrations were varied within the range optimized for each substrate, while the protein concentration was kept at 12 nM with respect to p53 tetramers. Under these conditions, the binding analysis showed that indeed three-stranded junctions were bound by p53 with a high affinity of 10⁻¹⁰ to 10⁻¹¹ M at a single binding site, as listed in Table 1 and illustrated in Fig. 5 by first-order binding curves and by the transformation of data into linear Scatchard plots. Perfectly matching oligonucleotides were bound comparably strong as was the substrate containing two G-T mismatches interrupted only by a single base pair, reflecting exactly the configuration generated in heteroduplexes with SV40-tsVP1(286S) DNA. Only minor deviations from these values were found with G-T or A-C mismatch substrates. In contrast, by measuring 12- to 15-fold-lower K_D values, we clearly discriminated interactions of p53 and three-stranded DNAs with the unpaired bases C-T or A-G, the mismatches produced by DNA exchange between SV40-tsVP1(290T) and SV40-tsVP1(196Y) genomes. Consistent with the finding of higher affinities toward A-G and C-T mismatch-containing substrates, we estimated faster on rates in these cases, as complete targeting of these DNAs at 100 pM was achieved within less than 2 min, whereas match substrates were maximally bound after 4 to 8 min (data not shown). Off rates appeared to be the same for both junction DNA types within our detection limits, and in agreement with data of Lee et al. (48), they indicated >80% substrate release within 1 min after addition of excess competitor DNA (data not shown).

TABLE 1.

Effect of mismatch type on the affinity of p53 to three-stranded DNA junctions

Mismatch type	Dissociation constant KD (10−12 M)
A-G	17 ± 8
C-T	21 ± 8
G-T	588 ± 138
A-C	99 ± 47
G-T/G-T	244 ± 66
G-C (match)	256 ± 113

FIG. 5.

Scatchard analysis of three-stranded match and A-G mismatch DNA binding. Three-stranded DNA junctions (see the legend to Fig. 3) consisted of SV40 sequences with mismatches introduced at the locus corresponding to VP1(P286S)/C2354. EMSA reactions were performed with constant p53 concentrations (12 nM for the tetramer) and increasing concentrations of radioactively labeled DNA substrate. Polyacrylamide gels were evaluated by PhosphorImager quantitation. Dissociation constants and standard errors were calculated from Scatchard plots, applying computational analysis by the program Grafit. (A) Graphic presentation of bound versus total substrate DNA results. (B) The dissociation constants calculated from the Scatchard plot for wild-type p53 are 256 × 10⁻¹² with match substrate and 17 × 10⁻¹² with A-G mismatch substrate. b/f, bound/free.

DNA damage in the form of ssDNA or mismatch is recognized by the C-terminal end of p53 (5, 37, 47, 63). To understand if binding of three-stranded match and mismatch DNAs is related to this damage-sensing function, we performed comparative competition studies with increasing concentrations of match and A-G mismatch DNA junctions, single-stranded top oligonucleotide, and the antibody Pab421 directed to the extreme C-terminal end (30, 34). Under conditions of equilibrium binding with saturating protein concentrations, labeled match and A-G mismatch substrates were completely bound and were competed equally well by unlabeled three-stranded match competitor (Fig. 6). Under the same conditions, mismatch competitor was slightly more effective, as can be deduced from the release of free probe at a concentration of 1 nM three-stranded A-G competitor and the disappearance of octameric p53-DNA complexes at 0.5 nM. Interestingly, we observed stabilization of protein-DNA complexes in the presence of single-stranded top oligonucleotide, as reflected by the constant proportion of double-tetramer forms, which became especially apparent with labeled match substrate and 120 nM competitor oligonucleotide. The three-stranded substrates could be displaced only at ssDNA concentrations above 120 nM. This experiment also showed that junction DNAs represent 300-fold-better competitors than ssDNAs. From this fact, we could exclude the possibility that the high affinity toward three-stranded match and mismatch DNAs can be explained simply by binding to the 25- to 37-nucleotide single-strand overhangs of the top and bottom oligonucleotides. Pab421 had a dual effect on the binding of artificial recombination intermediates: p53 tetramers were supershifted at concentrations equimolar and higher than the p53 tetramer input. However, the same antibody also competed with substrate binding, as seen best with mismatch-containing DNA.

FIG. 6.

Competitor studies for three-stranded match and A-G mismatch substrates. Radioactively labeled three-stranded junctions (see the legend to Fig. 3) without (A) or with (B) an A-G mispair at the position corresponding to the VP1(P286S)/C2354T locus (50 pM, final concentration) together with unlabeled specific competitors (3s or 3s A-G) at increasing concentrations from 0.5 to 8 nM were incubated for 25 min on ice with protein fractions enriched for wild-type p53 (wtp53) (25 nM tetramers). Unlabeled top oligonucleotide (1s) was used as the competitor at concentrations ranging from 1.2 to 12,000 nM, and the monoclonal antibody Pab421 directed to the C-terminal 30 amino acids was used at 3 to 300 nM.

DISCUSSION

Effect of p53her on homologous recombination.

In this study, we successfully applied our SV40-based recombination assay (89) to monkey cells specifically designed for the analysis of in vivo p53 functions in this process. By ectopic expression of the p53her protein at physiological levels, the growth-inhibitory effects described for wild-type p53 could be minimized before wild-type p53 induction by β-estradiol. p53her was functionally active, as demonstrated most clearly by strong and estradiol-dependent transcriptional transactivation via p53-specific DNA elements. At this point, we cannot exclude indirect effects of p53her-dependent growth delay or transcriptional transactivation on recombination processes in LLC-MK₂(p53her) cells. However, replication of the template, i.e., SV40 DNA, and production of the indicator, i.e., wild-type virus particles, displayed maxima at the same time points after infection of LLC-MK₂(p53her) and LLC-MK₂(neo) cells (data not shown). Most importantly in support of a more direct involvement in recombination, we collected data with the LLC-MK₂(p53her) cells and different tsVP1 variants, indicating that p53 inhibits interchromosomal homologous recombination events in a heteroduplex-dependent manner with evidence for mismatch-specific responses by p53.

p53 may have a role in early events in recombination.

Wild-type virus generation from two genomes mutated differently at a distance of 270 to 283 nucleotides as detected by our recombination assay system requires that the underlying mechanisms be complex events comprising either double crossover, two separate gene conversion events, or reciprocal exchange accompanied by a gene conversion event. Reciprocal exchange processes are less frequent than gene conversion events, with double crossovers being even more rare. From the proximity of recombination markers, coconversion events rather than independent correction by two gene conversion events affecting opposite DNA strands would be predicted (51). Favoring the third possible interpretation of DNA exchange between SV40-tsVP1 pairs, a mechanistic association between reciprocal and nonreciprocal exchanges was postulated (14). For yeast, the occurrence of coupled recombination processes was attributed to the effects of heteroduplex repair initiated by double-strand cutting at mismatches during the initial recombination event (31).

It is well established that after DSB formation the alignment of two homologous DNA molecules in register precedes the exchange process and that already at these early stages mitotic or meiotic checkpoint systems are involved in monitoring the status of recombination (91). On the basis of their studies of murine cells, Waldman and Liskay (83) favored a model whereby certain factors would scan the recombination intermediate for the minimal length of perfect homology to allow recombination to proceed further. In this context, it was also proposed that an initial branch migration process immediately after strand invasion would be very sensitive to mismatches. p53 has been linked physically (73) and genetically (49, 77) to Rad51, which performs the initial pairing and strand transfer steps in eukaryotes (10, 26, 74). In addition, we have demonstrated here that p53 recognizes artificial three-stranded DNA structures which allow some movement at the junction point very analogous to the situation in early recombination intermediates or in DNA joints during branch migration. We also accumulated evidence for an increased binding affinity in vitro, when p53 encounters distinct mismatches in these structures. Our measurements of 71-fold-lower recombination frequencies with the SV40-tsVP1(196Y)/SV40-tsB11 versus SV40-tsVP1(196Y)/SV40-tsVP1(290T) pair must be attributed to heterologies in the environment of the two phenotypically relevant loci. As wild-type p53 down-regulates recombination processes even further from this low level, it must play a regulatory role in a surveillance pathway, which cannot act upstream or as the downstream effector but rather acts in parallel to the postulated early mismatch scanning system responsible for the monitoring of uninterrupted homologies in heteroduplexes.

Extensive studies of sequence-specific DNA binding by p53 have led to the identification of numerous target genes, among which at least six seem to be relevant for the execution of cell cycle regulation and apoptosis after DNA damage (reviewed in reference 43). Analysis of the affinity for the p53 response element in the gene coding for the cyclin-dependent kinase inhibitor WAF1 with the bacterially expressed central core domain indicated a strong cooperativity for binding of four molecules and the importance of bending or flexibility in the DNA substrate (8). The K_D value of 8 × 10⁻⁸ for the p53-WAF1 promoter interaction is lower than those determined for other transcription factors (41, 78), although for the whole versus the truncated p53 molecule, one must expect an enhancement due to the contribution of the oligomerization domain and to activating p53 modifications (32, 35). From the data presented here, it becomes clear that p53 is targeted to recombination intermediate-like DNA extremely specifically (K_D = 10⁻¹⁰ to 10⁻¹¹ M), implying that important functions are performed by p53 during DSB repair and during mitotic and meiotic recombination. Intriguingly, mutant p53 binds MAR/SAR DNA elements with a similar affinity (86), and again this type of DNA recognition seems to be coupled to the structural characteristics of bends (58). It has been reported that p53 adopts a mutant-like conformation upon DNA binding (29), but the supershifts shown here indicate that p53 remains wild type with respect to Pab246 immunoreactivity when bound to three-stranded DNAs. Although mismatches contribute to localized flexibilities, which is an important factor for DNA interactions of proteins actively bending DNA (25), we consider this unlikely to be the sole explanation for the increased affinities to mismatch-comprising three-stranded DNAs, since the same unpaired region in flayed duplexes failed as a substrate and since preferences became apparent only for distinct base-base mismatches. In accordance with our data, Lee and colleagues (47) presented evidence for binding of unpaired bases by p53 and interpreted this activity as important for DNA damage sensing. Mismatch recognition within recombination intermediates might implicate the active participation in surveillance mechanisms.

Possible role of mismatch binding activity of p53 in minimizing homologous recombination.

Here, we provide evidence that mismatch specificities of p53 correlate with recombination-regulating activities. It remains unknown whether and how the exchange process is attenuated when p53 encounters A-G or C-T nucleotide in the strands aligned in heteroduplexes. From the exonuclease activity intrinsic to p53 (59), one can envision degradation and repair of the unpaired site within the intermediate, as performed by bacterial exonucleases during postreplicative mismatch repair initiated by the MutS/L/H system (reviewed in references 44 and 57). Mismatch repair activities on recombination intermediates are known for the yeast mismatch repair factors MSH2, MLH1, and PMS1, thereby causing gene conversion. In contrast, we find p53 to inhibit recombination processes which most probably comprise gene conversion. In our assay system, we detect successful recombination events by wild-type SV40 production from two mutant templates identical except for two to three base pairs. From our data obtained with cells lacking functional wild-type p53, preferential conversion of the mutant information to wild-type information in the case of SV40-tsVP1(196Y)/SV40-tsVP1(290T) compared to SV40-tsVP1(196Y)/SV40-tsVP1(286S) pairs would imply mismatch-type-specific repair acting upon the mispairing at 290T. If this were true for our assay setup, p53 would have to inhibit elevated gene conversion at the 290T locus by suppressing the preferential excision of mutant T opposite wild-type C nucleotides or A opposite G. Preferential recognition of A-G or C-T versus G-T or A-C single-base mispairings has not yet been described for human cells with respect to either MSH2-dependent or MSH2-independent binding (62).

Alternatively, according to the DSB repair model (75) of genetic recombination, unilateral DNA exchange, as detected here by wild-type virus production, could arise from degradation of the mutant strand followed by recombination repair DNA synthesis of the gap with wild-type DNA as the template. In this scenario, p53 would have to either inhibit nucleolytic degradation of the mutant strand or promote degradation of wild-type DNA, suppress initiation of recombination by strand invasion, inhibit repair replication, or cause a bias in repair replication toward the mutated sequence. Further experiments are necessary to determine whether p53’s exonucleolytic activity plays a role in suppressing recombination. To distinguish between the remaining possible mechanisms underlying recombination inhibition by p53, we consider it crucial that our in vitro DNA binding data demonstrate high specificities for three-stranded junctions compared to those for double-stranded fork structures emerging during replication, which is the reason why we favor a role of mismatch recognition by p53 during the early strand invasion process.

In addition to their well-known role in repair, mismatch repair components were also described to prevent the detrimental effects of genetic exchange between divergent sequences by inhibiting the completion of recombination (reviewed in references 44 and 57). Some ideas on the mechanism underlying this type of inhibition of homologous strand transfer can be drawn from in vitro experiments with bacterial MutS. MutS binds Holliday junctions and inhibits RecA-induced strand exchange and branch migration upon encountering mismatches in the heteroduplex. In analogy, it is conceivable that p53’s interaction with Rad51 (73) serves to enzymatically inactivate Rad51 and that this is triggered or enhanced by high-affinity mismatch recognition in the nascent heteroduplex. Alternatively, p53 could be envisioned to either recruit destabilizing helicases, such as XPB and XPD (84, 90), or actively dissolve intermediates comprising A-G or C-T mispairs. Thus, by promoting reverse branch migration via its reannealing or strand transfer activity (4, 15, 61), p53 might attempt to reconstitute the matching templates, a subject of our further investigations. It is interesting that p53 binding to mismatch-containing three-stranded intermediates is sensitive to competition by the antibody Pab421, the antibody known to inhibit strand transfer (63).

High affinities accompanied by fast on and off rates for p53 interactions with three-stranded structures point to a dynamic process, such as reiterative binding within a homology scanning mechanism. Stabilization of these protein-DNA complexes might be achieved under certain conditions in vivo, for which multimerization would be a good candidate. p53 was described to form double tetramers by interactions via the central domain, a mechanism suggested to result in DNA loop formation (72). In comparison to tetramerization via the oligomerization domain, double-tetramer formation might require protein units within the tetramers not bound to DNA internally, as competition with three-stranded junctions causes disappearance of double versus single tetramers at four- to eightfold-lower concentrations. In competition experiments with complementary ssDNA, we observed p53 to stay in both DNA-bound tetrameric forms, as long as the concentrations of the competitor oligonucleotide were below or approximating the concentrations of p53 monomers in the solution. We believe that interaction of p53 with ssDNAs, presumably via its C terminus (5), stabilizes the DNA-bound protein oligomers cooperatively. This phenomenon might be related to the stimulatory effect of single-stranded oligonucleotides on sequence-specific binding by p53 (37). Therefore, similar to the role of PMS2 and MLH1 during chromosome synapsis in meiosis (6, 7, 22), p53 as part of double-tetramer complexes might stabilize and examine nascent heteroduplexes, thereby causing a delay of homologous recombination processes in general and blockage of homologous exchange upon encountering mismatches.