and were markedly more abundant in an immortal cell line than in the diploid normal cells from which that ...... Thomas, K. R., K. R. Folger, and M. R. Capecchi.
Vol. 9, No. 9
MOLECULAR AND CELLULAR BIOLOGY, Sept. 1989, p. 4009-4017 0270-7306/89/094009-09$02.00/0 Copyright © 1989. American Society for Microbiology
Homologous Plasmid Recombination Is Elevated in Immortally Transformed Cells GREGORY K. FINN, BENEDIKT W. KURZ,t RICHARD Z. CHENG, AND ROBERT J. SHMOOKLER REIS* Departments of Medicine and Biochemistry and Moleciulair Biology, Unih'ersity of Arklansas for Medical Sciences, and Geriatric Research, Education, and Clinical Ceniter, Medical Research/ISI, McClellan Veterans Admninistration Medical Center, 4300 West 7th Street, Little Rock, Arkansas 72205 Received 9 January 1989/Accepted 13 June 1989
The levels of intramolecular plasmid recombination, following transfection of a plasmid substrate for homologous recombination into normal and immortally transformed cells, have been examined by two independent assays. In the first assay, recovered plasmid was tested for DNA rearrangements which regenerate a functional neomycin resistance gene from two overlapping fragments. Following transformation of bacteria, frequencies of recombinationlike events were determined from the ratio of neomycin-resistant (recombinant) colonies to ampicillin-resistant colonies (indicating total plasmid recovery). Such events, yielding predominantly deletions between the directly repeated sequences, were substantially more frequent in five immortal cell lines than in any of three normal diploid cell strains tested. Effects of plasmid replication or interaction with T antigen and of bacterially mediated rejoining of linear molecules generated in mammalian cells were excluded by appropriate controls. The second assay used limited coamplification of a control segment of plasmid DNA, and of the predicted recombinant DNA region, primed by two sets of flanking oligonucleotides. Each amplified band was quantitated by reference to a near-linear standard curve generated concurrently, and recombination frequencies were determined from the ratio of recombinant/control DNA regions. The results confirmed that recombinant DNA structures were generated within human cells at direct repeats in the transfected plasmid and were markedly more abundant in an immortal cell line than in the diploid normal cells from which that line was derived.
mutations, in carcinogenesis (6, 30, 51, 53), a view supported by the remarkably large fraction of spontaneous somatic mutations which are DNA rearrangements, primarily deletions (9). As a first step in evaluating the role of recombination in cell transformation to unlimited replicative potential (immortality), we have compared rates of extrachromosomal plasmid recombination in normal diploid versus immortally transformed aneuploid cells. We report here a striking increase in immortal cell lines, relative to vigorous normal diploid cell strains, of plasmid rearrangements with features of homologous recombination.
Genetic heterogeneity is one of the most remarkable and pervasive characteristics of established or immortal cell lines, especially those of high tumorigenicity (6, 32, 33, 53). Whereas normal cells in vivo or in culture are predominantly or entirely diploid, tumors and immortal cell lines are invariably aneuploid. Such transformed cells exhibit frequent chromosomal alterations, in particular translocations (53) and gene amplifications (6, 37), some of which are associated with specific tumor types (6, 24, 53). They also undergo progressive dedifferentiative changes in gene expression, such as ectopic production of gene products (typically embryonal genes) which are not found in normal differentiated cells of that type (32), and clonal loss of gene products which would be expressed in normal cells of that type (33). These factors imply a fundamental shift toward genetic instability associated with transformation to immortality and tumorigenesis. Although homologous recombination is a likely common mechanism for DNA translocation and amplification (1, 10, 17) and a possible mechanism for alterations in gene expression (6, 24), recombination frequencies have not been directly compared between immortally transformed cell lines and normal diploid cells. Most carcinogens prove to be mutagenic when assayed in Salmonella His- reversion assays which are specific for point or frameshift mutations (6), but many are poorly detected by such mutation assays, resulting in concordances of -60% (23, 46). It has been argued that DNA rearrangements are strongly implicated, perhaps even more than point
MATERIALS AND METHODS Human diploid fibroblast strains. MRC5 cells were derived from normal human fetal lung; strains A2 and J089 were obtained from forearm biopsies of male donors, ages 11 and 68, respectively. These strains have replicative life spans in our laboratory of approximately 70, 65, and 44 mean population doublings, respectively. Immortally transformed human cell lines. MRC5-SV cells were derived by simian virus 40 (SV40) transformation of MRC5 fibroblasts. HeLa-A and HeLa-B cells are subclones of the HeLa cervical carcinoma cell line, presumed to have been transformed by papillomavirus (8). CACL-73-36 cells were obtained from a human myeloma; 293 cells are adenovirus-transformed human kidney cells. XPA-SV (GM 5509; National Institute of General Medical Sciences Human Genetic Mutant Cell Repository, Bethesda, Md.) cells are SV40-transformed fibroblasts characterized as Xeroderma pigmnentosium complementation group A. Transfection of plasmid DNA into mammalian cells. Plasmid DR1 (34) (see Fig. 1) was grown in DH5 Eschzerichia coli
* Corresponding author. t Present address: Washington University Medical School, St. Louis, MO 63110. v Present address: Department of Microbiology. University of Arkansas, Fayetteville, AR 72701. 4009
4010
FINN ET AL.
and recovered by conventional procedures. Cell lines and strains were maintained at 37°C under 5% CO2 in Eagle minimal essential medium supplemented with 15% fetal calf serum. Cells were grown to approximately 95% confluence for transfection by DEAE-dextran with dimethyl sulfoxide shock and were thereafter treated as described by Lopata et al. (29), using 5 ,ug of plasmid DNA per 100-mm dish of cells. Plasmid DNA recovery from mammalian cells. Plasmid was extracted from infected cells by the method of Hirt (19), with the following additional purification steps (S. Subramani, personal communication). The genomic Hirt supernatant, after centrifugation for 15 min at 46,000 x g at 0°C, was recovered and centrifuged again. Each supernatant (-0.9 ml) was mixed with 100 ,ul of 3% sodium dodecyl sulfate0.25 M EDTA-0.5 m Tris hydrochloride (pH 9.5) and heated for 15 min at 65°C. Samples were cooled on ice and mixed with 0.2 ml of potassium acetate (pH 4.8). After 1 h at 0°C, they were centrifuged as described above. Samples were extracted twice with 1:1 phenol-chloroform and twice with 24:1 chloroform-octanol. Aqueous phases were precipitated with 1 volume of isopropanol at -20°C for 16 h and centrifuged as described above. Pellets were suspended in 450 Rl of 1 x TE (10 mM Tris hydrochloride, 1 mM EDTA [pH 8.3]) and incubated for 15 min at 37°C with 20 ,ug of RNase A per ml. LiCl was added to 0.8 M, and samples were precipitated with 2 volumes of ethanol at -20°C for 16 h and centrifuged as described above. Dried pellets (each representing one to two dishes of cells) were resuspended in 100 [ul of lx TE. Quantitation of recovered plasmid DNA. Both undigested and EcoRI-digested samples of Hirt extracts (as described above) were run on 1% agarose gels. Plasmid recovery was quantitated by ethidium bromide fluorescence of bands in gels or by autoradiography after alkaline blotting (31) onto nylon membranes (Zeta-Probe; Bio-Rad Laboratories) and hybridization (13) to 32P-pBR322 probe, homologous to 2.3 kilobase pairs (kbp) of DR1 (PvuII-EcoRI). Bacterial assay of plasmid recombination. Low-molecularweight DNA recovered from mammalian cells and from plasmid DNA controls was used for transformation of competent DH5 bacteria (18). Competent cells were stored at -70°C for up to 8 months with no appreciable loss in efficiency, which ranged from 0.4 x 109 to 1.1 x 109 colonies per ,ug of pUC19 plasmid. Recombination assays used 2 to 16 ng of recovered plasmid DNA per 400 [1I of bacteria (>2 x 108 competent cells), which was then concentrated by centrifugation and suspension in SOC medium (18) before being plated onto ampicillin and neomycin plates. Plasmid recombination assay by polymerase chain reaction, using primers flanking the homologous region. The in vitro amplification by TaqI DNA polymerase of DNA sequences primed with flanking synthetic oligonucleotides (polymerase chain reaction, or PCR) has been performed essentially as described by Saiki et al. (36). Two oligonucleotide primers (NP28 and NP30') were synthesized, flanking the overlap region of the gene encoding neomycin resistance (Neor), and two control primers (CP24 and CP25') were synthesized, situated 300 bp apart near the pBR322 ori (see Fig. la and b). The sequences of the primers (and their positions with respect to the GenBank sequences from which they were obtained [TNSNEO for NP28 and NP30' and pBR322 for CP24 and CP25']) are as follows: NP28 (328 to 355): NP30' (818 to 789): CP24 (3001 to 3024): CP25' (3348 to 3324):
5'-CTGCAGGACGAGGCAGCGCGGCTATCGT-3' 5'-AACGCTATGTCCTGATAGCGGTCCGCCACA-3' 5'-GCTGAAGCCAGTTACCTTCGGAAA-3' 5'-GAACGAAATAGACAGATCGCTGAGA-3'
MOL. CELL. BIOL.
Amplification consists of 12 to 20 synthesis cycles with the Thermus aquaticus DNA polymerase (Perkin-Elmer/Cetus) and either 1 to 500 pg of plasmid DNA or 0.5 to 1 pul of Hirt supernatant (containing 0.1 to 10 ng of DNA), in 50 ,ul of buffer containing 50 mM KCl, 10 mM Tris hydrochloride (pH 8.4), 2.5 mM MgCl2, 0.25 or 1 puM of each primer, 0.2 mM of each deoxynucleoside triphosphate, 0.1 mg of gelatin per ml, and 1 U of polymerase (36). Each cycle, for these primers, comprises 1 min at 93°C (denaturation), 3 min at 45°C (reannealing), and 3 min at 72°C (polymerase extension), controlled by an electronic thermal cycler of our own design (R. Z. Cheng et al., manuscript in preparation). The DNA is then directly analyzed by agarose gel electrophoresis and quantitated by densitometric scanning of autoradiographs, after Southern blot hybridization with [32P]DNA Neor gene probe. The presence of a sharp, Neor gene-hybridizing band at 0.49 kbp indicates the presence of recombinant molecules as shown (* indicates a HaeIII site): AnOAQk....- I
PstI
*
-
NP30'
SinaI
Background amplification can be kept extremely low in the TaqI polymerase reactions, by the use of GC-rich primers larger than 24 nucleotides coupled with an appropriately high annealing temperature. Statistics. Differences between means are tested for significance level by the Behrens-Fisher test, a stringent version of the Student t test appropriate to comparisons of samples with any n c 10 (20). Significance levels are indicated as either P2 for a two-tail t test, or P1 for a single-tail integral, for which the sign of X1-X2 has been determined. RESULTS Transient assay of plasmid recombinant frequency. In order to compare transformed immortal cells with diploid cells of limited replicative potential, homology-dependent DNA rearrangements were assessed for plasmid markers introduced in a transient assay. Cells were transfected with plasmid DR1 (34), which contains two incomplete but overlapping fragments of the Neor gene separated by a 2.9-kbp Escherichia coli gpt gene (Fig. 1). The 5' and 3' Neor gene fragments, in the same orientation, share a 420-bp region of overlap such that homologous recombination between the duplicated segments would produce a single, complete neomycin resistance gene. Plasmid recombination thus refers here to the generation of functional (recombinant) Neor genes from two nonfunctional overlapping gene fragments and is not intended to imply either a mechanism or random exchange of flanking markers (see Discussion). The DR1 plasmid also contains a functional gene encoding ampicillin resistance (Ampr), external to the potential recombination region and thus not deleted during intramolecular homologous recombination, and an SV40 ori-promoter region just 5' to the split Neor gene (Fig. la). The assay of plasmid recombination was essentially as described for COS cells by Rubnitz and Subramani (34) and by Ayares et al. (4). Closed-circular plasmid DNA was introduced into mammalian cells, using DEAE-dextran with dimethyl sulfoxide shock (4, 29, 44) because of its very high uptake efficiency (29, 44), greater reproducibility (29), and substantially better recovery of Ampr colonies (5) than obtained after transfection by calcium phosphate coprecipitates. Extrachromosomal DNA was recovered from the mammalian cells 12 to 72 h after transfection, using a
VOL. 9, 1989
PLASMID RECOMBINATION IN NORMAL AND TRANSFORMED CELLS
4011
A,
zco-
CPT
DR3 deletion
(9 - NdeI
+ - EcoRI
?
+-
-
Sf ii
HindlIl
FIG. 1. Map of plasmid DR1, showing overlapping Neor gene fragments. (a) Schematic map of DR1 (redrawn from reference 34). The pBR322 region contains the plasmid origin of replication and Ampr gene. NEO-1 includes the SV40 ori-early promoter (SV-ORI), the 5' section of the TN5 Neor gene (A) containing its long terminal repeat-promoter region (_), and the 420-bp overlap segment (B); NEO-2 contains segment B repeated in the same orientation and the 3' section of Neor (C). The 2.9-kbp E. coli gpt gene, inserted between NEO-1 and NEO-2, would be deleted by intramolecular homologous recombination between the B segments, as illustrated in panel b.
modification of the procedure by Hirt (19). Plasmids were tested for generation of recombinant Neor genes in E. coli DH5, which is more stringently RecA- than HB101 cells (18), thus minimizing the background of bacterial recombination. The DH5 cells were made competent (18), with transformation efficiencies of 0.4 x 109 to 1.1 x 109, monitored by using pUC19. After transformation with pure plasmid or low-molecular-weight recovered DNA, each sample of bacteria was then scored for Neor on neomycin plates and for Ampr on ampicillin plates. Because plasmid recombination rates were calculated from the ratio of recombined (Neo9) to total (Amp) plasmids for each cell line and time point, results were internally controlled for the recovery of viable plasmid. Assays showed good agreement across a 10to >40-fold range of Hirt supernatant or plasmid DNA inputs (Fig. 2A and B illustrate colony yields and Neor/Ampr ratios for three DNA inputs in a single experiment). After transformation of DH5 E. coli with DNA extracted from DR1-transfected MRC5 and MRC5-SV cells, plasmids were analyzed from six neomycin-resistant clones each. All recovered plasmid clones (12 of 12) were 5-kbp monomers, since uncut DNAs comigrated with 5-kbp circular DNA standards. Moreover, all yielded identical restriction maps consistent with homologous recombination leading to excision of the E. coli gpt gene (data not shown). These results suggest that the viable Neor plasmids recovered in this assay were formed predominantly by deletion of DNA between direct repeats. Plasmid recombination frequency as a function of cell transformation to immortality. The fraction of recombinant plasmids increased with time after transfection into an SV40-transformed cell line (MRC5-SV), with a maximum at 36 to 48 h, but remained quite low and stable over time in the MRC5 diploid strain from which MRC5-SV cells were derived (Fig. 2C). The proportion of recombinant (Neor) plasmids within extrachromosomal DNA increased by 10- to 20-fold in SV40-transformed MRC5 cells, compared with that in normal MRC5 cells, at 36 and 48 h posttransfection
(after subtraction of E. coli background; Table 1). The differences between recombination rates for MRC5 versus MRC5-SV at 36 and 48 h posttransfection (6 to 13 assays each) were statistically quite significant, at P2 < 10-6 (20). Two additional diploid fibroblast strains, from normal donors, were examined at 24 to 48 h after transfection with DR1 plasmid. Both strains yielded low levels of plasmid recombination, comparable to the results for normal MRC5 cells (Table 1). Four additional immortally transformed human cell lines, tested at 36 h posttransfection, supported recombination frequencies ranging from 1.5- to 4.5-fold greater than the highest normal cell value (after background subtraction; Table 1). Each of the five immortal cell lines tested was significantly higher in plasmid recombination than any of the three normal control cell strains (P1 < 0.025 to P1 < 10-6 for each pairwise comparison, except for J08924h versus XPA-SV [P1 < 0.05] or CACL [not significant]). Plasmid recombination is not due to bacterial religation of DNA molecules linearized in mammalian cells. When extrachromosomal plasmid DNA recovered from transfected human cells was electrophoresed without enzyme digestion and was examined by hybridization of Southern blots to a DR1-specific probe, the resulting autoradiographs (Fig. 3) indicated extensive nicking of DR1 DNA soon after its introduction into mammalian cells. Although plasmids were >98% supercoiled at transfection, conversion to the nicked circular form was essentially complete by 12 to 24 h in human cells, whether diploid or SV40 transformed, with linear molecules constituting an approximately constant fraction (5 to 10%) between 12 and 48 h posttransfection. We were concerned that linear molecules generated after transfection might then serve as preferred substrates for bacterial recombination or rejoining or both, especially in view of the relatively high frequency reported for such events (34). However, the extent of nicking or linearization of plasmid during its residence in mammalian host cells was not correlated with recombination frequency in this assay (Table 1; Fig. 3). In order to test this point conclusively, in
4012
MOL. CELL. BIOL.
FINN ET AL.
0
250.
z
Po
zt , '
200
0
B
0.
0 3
t0 "I-
-80.000
150
~
O
.
MRC5amp A
0
aL) 100 MRC5-SVamp A or g3 Oo 50 MRC5-SVneo 54
-20,000
100
0
a)
S
MRC5-SV ------
-80.000 0
,'
..0
4
150-
*~
50-
MRC5 SB
; 0
0.1
MRC5neoOG
o 0.1
PeA.
10
1
Input DNA (ng) FIG. 2. Plasmid recombination, assessed by bacterial transformation, after transfection of DR1 plasmid into diploid MRC5 cells or immortally transformed MRC5-SV cells. Neor and Ampr colony yields (A) are roughly proportional to input of low-molecular-weight DNA, resulting in relatively stable Neor/Ampr colony ratios (B). (C) Time course of plasmid recombination. Values are means of duplicate assays, after subtraction of background (1.8 x 10-5) caused by residual bacterial recombination (DH5 E. coli transformed by pure DR1 plasmid). Duplicate transfections gave qualitatively similar time courses, with MRC5-SV values 20- to 33-fold higher than MRC5 at 1.5 and 2 days but only 3- to 7-fold higher at 0.5 and 1 day MRC5-SV cells. posttransfection. Symbols: _, MRC5 cells;
10
1
Input DNA (ng)
d,_ 12'0O C .° 0
L0 O
e0
,
X
l
2
Q
0.0
0.5
1.0
1.5
2.0
Days after Transfection several experiments the linear molecules were specifically removed from extracts (2, 52) by exhaustive digestion with ATP-dependent exonuclease (U.S. Biochemical Corp.) (Fig. 4A). In each case, the substantial difference between SV40transformed and normal MRC5 fibroblasts was maintained, with no significant change in recombination rates (Fig. 4B). Marker recombination is independent of plasmid replication or interaction with T antigen. High levels of T antigen, as
observed in COS African green monkey cells, can allow the replication of a virus or plasmid containing an intact SV40 ori region (16). Previous studies with SV40 implied that neither T-antigen expression nor viral replication is required for homologous recombination in monkey cells (21, 42), but replicating molecules have been reported to have either elevated (4, 21) or depressed (49) recombinogenic potential. It was therefore possible that a large increase in plasmid replication within MRC5-SV cells caused by T-antigen TABLE 1. Recombination frequencies in normal and immortally transformed human cellsa Cell strain
a
b
c
d
e
f
.-@~1 &^ is
h
g
ij
anc
Normal fibroblasts MRC5d MRC5d A2 A2d J089d J089d Transformed cells MRC5-SVd MRC5-SVd HeLa-A HeLa-B 293 XPA-SV CACL E. colie d
..
a ..
< CCCD'
*
FIG. 3. Analysis of DR1 plasmid in low-molecular-weight DNA extracts from MRC5 and MRC5-SV cells 12 to 48 h after transfection. Samples of Hirt supernatants (each representing 1 x 105 to 2 x 105 cells transfected with 5 jig of DR1 DNA per 10-cm dish) were electrophoresed in 1.5% agarose gel for 16 h at 2.5 V/cm and transferred in alkali to a nylon membrane. Hybridization was with [32P]DNA of plasmid pBR322, followed by stringent washing and autoradiography (8 h). Lanes: a to c, Hirt samples from MRC5 diploid cells at 12, 24, and 48 h posttransfection, respectively; d, 100 ng each of linear (Ndel-cut) and uncut DR1 as markers; e to g, Hirt samples from MRC5-SV transformed cells at 12, 24, and 48 h, respectively; h, 60 ng of Ndel-linearized DR1; i, undigested DR1; j. 60 ng of uncut pBR322. nc, Nicked circular DR1; ccc, covalently closed circular DR1; lin, linear DR1 molecules.
Time
(h)b
10-5
frequency (n)'
36 48 24 48 24 36
12.4 9.4 7.4 13.6 15.0 12.0
36 48 36 36 36 36 36
129 149 55.4 46.2 60.9 28.0 21.7 1.9
± ±
± ± ± ±
± ± ±
±
±
2.4 1.6 1.5 2.1 5.3 2.3
(10) (6) (4) (4) (4) (4)
6.3 (13) 9.6 (9) 14.5 (8) 11.0 (8) 10.1 (9) 2.2 (8) 3.1 (9) 0.18 (23)
" DR1 plasmid was introduced into all cell strains and lines indicated, by the DEAE-dextran-dimethyl sulfoxide transient transfection procedure (27, 43), and recombination was assessed by transformation of DH5 bacteria to Neor or Ampr (see Materials and Methods) at various times after transfection. Times are posttransfection intervals. Recombination frequencies were obtained as the ratios of Neor/Ampr colonies in replicate platings and are given as the means + standard error of the mean of values obtained in n independent transformation assays. " Time courses were followed as for Fig. 2C, and the data for these cells were at or near the maximal yield of recombinants. Other values (for four immortal cell lines) may be below maximal level. ' DR1, not transfected into mammalian cells.
PLASMID RECOMBINATION IN NORMAL AND TRANSFORMED CELLS
VOL. 9, 1989
a
b
c
d
e
f
5,2>*j
h
g
j
i
k
m
n
4013 o
q
p
;i
I.8>
.~-
0
0
cd04
o.3D
0
0.12
0.5
2.0
B Input of ATP-dependent DNAse (units) FIG. 4. Selective digestion of linear DNA molecules by ATPdependent exonuclease. (A) Each sample contained 100 ng of uncut DR1 and 30 ng of NdeI-linearized DR1 DNA. Samples also contained 10 ,ul (10% of one 10-cm dish, or 1 x 105 to 2 x 105 cells) of Hirt supernatant either from MRC5 normal fibroblasts at 48 h posttransfection (lanes a to e) or from MRC5-SV transformed cells (lanes f to j). Lanes contained samples digested with the following amounts of enzyme (in units): a and f, none; b and g, 0.03; c and h, 0.125; d and i, 0.5; e and j, 2.0. Lanes at right are size standards. Abbreviations are as defined for Fig. 3. (B) Hirt supernatants, digested as described above but without added DR1, were tested for DR1 Neor region recombination by bacterial transformation. Error bars are standard error of the mean for replicate assays. Addition of 100 ng of PM2 DNA to all samples had no effect. ±
expression might contribute to their elevated recombination frequency. We have tested this possibility in three ways, as follows, with uniformly negative results. (i) Replication of plasmid DNA. Replicative synthesis of plasmid DNA in mammalian cells can be detected by the appearance (after two or more replication rounds) of molecules lacking the dam-dependent adenosine methylations initially present in the plasmid. Unmethylated GATC sites are cleavable by MboI but not by DpnI, in contrast to input plasmid (methylated at these sites), which was cleavable by DpnI rather than MboI. Detection of a small fraction of replicated, DpnI-resistant DNA could be compromised, however, by incomplete digestion of fully methylated molecules (Fig. 5, lane b), whereas plasmid sensitivity to MboI cleavage unambiguously demonstrates DNA molecules which have replicated in a host cell lacking the dam modification system. By the latter method, no resynthesis of the pBR322 sequences in DR1 could be seen for any cells tested, A2, MRC5, or MRC5-SV (Fig. 5, lanes d to g and j to m). The absence of plasmid DNA synthesis in MRC5-SV cells was probably due to down regulation of T-antigen expression in human cells transformed by complete SV40 (16, 21). (ii) T-antigen binding at the SV40 ori. COS1 cells were transfected with DR3, a plasmid we constructed from DR1 by deletion of a 64-bp SfiI-HindIII fragment (SV40 nucleotides 5171 to 5234; Fig. la) overlapping the origin of repli-
FIG. 5. Maintenance of dam site methylations in transfected DR1 DNA. Hirt supernatant samples, each representing _105 cells at 1 to 2 days posttransfection, were digested for 1 h at 370C with 2 U of DpnI or MboI enzyme, as indicated below. They were electrophoresed for 16 h at 2 V/cm in 1.5% agarose, transferred to nitrocellulose, and hybridized to a pBR322 [32P]DNA probe at 0.5 x 106 cpm/ml (lanes a to m) or DR1 probe (lanes n to q). Lanes: a to c, 0.1 ,ug each of uncut (a), DpnI-digested (b), or MboI-digested (c) DR1 without transfection; d to g, Hirt extracts from two preparations of uncut (d and f) or MboI-digested (e and g) DR1-transfected A2 normal fibroblasts; h and i, lighter exposures of lanes a and b; j and k, Hirt extracts from uncut (j) or MboI-digested (k) DR1transfected MRC5-SV cells; and m, Hirt extracts from uncut (1) or MboI-digested (m) DR1-transfected MRC5 normal fibroblasts; n to q, same as lanes j to m, dehybridized and reprobed with 32P-DR1, which also detects free SV40 molecules in MRC5-SV DNA samples (5.2 kbp, undigested, lane n; 0.5 and 1.9 kbp after MboI digestion, lane o). DNA fragment sizes (in kilobase pairs) are shown at the left.
cation and two of the three T-antigen-binding sites (16). If T-antigen binding to DR1 contributed in any way to its recombination, then DR3 should undergo less Neor gene recombination than DR1. Instead, we observed severalfoldhigher recombination for DR3 than for DR1 (compare lines 1 and 2, Table 2). No significant differences between DR3 and DR1 recombination were found in cell strains or lines other than COS1 (data not shown). Our data thus suggest that T-antigen binding at the SV40 ori region is somewhat inhibitory to homologous recombination monitored at the Neor overlap segment. TABLE 2. Recombination rates in primate cell lines: effect of cell type, plasmid, and method of transfection Cell line
1. COSi 2. COS1 3. CV-1
Plasmid
DR3 DR1 DR1
i0' Recombination frequency (colony count)a on day: 1
2
97 (3) 30 (15)
49 (10) 19 (29) 17 (16)
' Recombination frequencies shown are the means of duplicate assays, which differed by - 3 -, 4 -j
*_- 1
v/
,/ '~~~~ z~~~~~
!
/
I
,1 1
-3
-1
-2
0
Log [DNA] (ng) -0
c 60
C
_ 50-
_ 40-
/;
30* 20
7
/ i /
/
r -. r, 0--_iS
0.5
1.0
1.5
2.0
Days After Transfection
FIG. 6. Assay of recombinant Neor DNA regions by flankingprimer amplification (PCR) of recovered low-molecular-weight DNA. (A) Resolution by electrophoresis in 1% agarose gel of DNA products after 20-cycle TaqI polymerase amplification (PCR) with primers flanking the Neor gene overlap region (producing a 0.49-kbp band), primers flanking a control region of unduplicated pBR322 DNA (producing a 0.35-kbp band), or both. Samples were digested with excess HindIll (5 U/0.2 ng) prior to amplification to linearize DR1 or derived recombinant plasmids. Lanes: a, 100 ng of PM2HindlIl marker fragments (5.6, 2.1, 1.0, 0.46, 0.44, and 0.23 kbp); b to d, DR1 control DNA, not transfected; e to h, low-molecularweight DNA (Hirt supernatants) from MRC5 cells 36 h posttransfection (0.2 ng, initial DNA content); i to k, low-molecular-weight DNA from MRC5-SV cells 36 h posttransfection (0.2 ng, initial DNA content). Samples b, e, h, and i contained 1 ,umol of each Neor region primer per reaction mixture; samples c, f, and j contained 1 p.mol of each control region primer; and samples d, g, and k contained 1 ,umol of each Neor region primer and 0.25 ,umol of each control primer. Efficiency of the control-primed amplification is thus attenuated relative to the Neor region-primed amplification by reduction of input, CG content, and length of control primers.
(iii) Nonspecific effects of T antigen. COS cells, which are unusual in their high-level expression of T antigen, were derived by transformation of CV-1, a permanent line of African green monkey cells, with origin-defective SV40 (16). Since both lines are immortal, comparison of COS1 cells with their parental CV-1 line should indicate whether high T-antigen production itself affects the recombination rate. No significant difference was observed between COS1 and CV1 cells 2 days after transfection with DR1 (Table 2, compare lines 2 and 3). Thus, recombination within plasmid DR1 does not appear to depend on T-antigen expression by host cells. Quantitative analysis by PCR of elevated plasmid recombination in SV40-transformed human fibroblasts. We have examined the Neor gene overlap region of putative recombination, in low-molecular-weight DNA isolated from DR1transfected mammalian cells, to determine whether DNA rejoining was completed within human cells and whether there was precise alignment of the Neor gene direct repeats in recombinant plasmids. At the low frequencies of recombination observed in these cells, the restriction fragments expected from recombined plasmid could not be unambiguously resolved, by Southern blot analysis, from background signal caused by degradation of unrecombined DNA present in 103 to 105 excess. The region spanned by the DR1 direct repeats was therefore amplified by PCR and then analyzed by agarose gel electrophoresis and Southern blot hybridization. We synthesized primers NP28 and NP30', flanking the Neor gene direct repeats, which are separated by 3.7 kbp in unrecombined plasmid DR1 (Fig. la) but span only 0.49 kbp after homologous recombination (Fig. lb and diagram in Materials and Methods). A second pair of primers, CP24 and CP25', was included in reaction mixtures to prime the synthesis of an unduplicated 0.35-kbp region of DR1 (Fig. 1), which thus provides an internal control for plasmid recovery and efficiency of amplification. Figure 6A illustrates the results of adding these primer pairs, separately or combined, to plasmid DR1 (lanes b to d) and to Hirt supernatants from DR1-transfected MRC5 (lanes e to h) and MRC5-SV (lanes i to k) cells. These data indicate that low-molecular-weight DNAs from DR1-transfected mammalian cells indeed contain recombinant molecules, as defined by the amplification of sharp bands at the predicted size of 0.49 kbp, in far greater numbers than were present in the input plasmid. The 0.49kbp bands, confirmed by Southern blot hybridizations (data not shown) to correspond to the Neor gene direct repeat, were more abundant in extrachromosomal DNA recovered
(B) Standard curves generated by PCR amplification of various inputs (2 to 500 pg) of plasmid pSV2neo DNA. Replicate samples, amplified for 13 cycles, were fractionated on agarose gels, transferred to charged nylon filters, and hybridized to [32P]DNA of the Neor gene overlap region (B in Fig. la). Autoradiographic signals for various times were quantitated with a scanning densitometer (model GS300, Hoeffer Scientific Instruments) and corrected for exposure and 32P decay. A log-log slope of 1 (indicated by thin dashed lines), which fits data up to 50 to 100 pg, indicates a first-order or linear relation between input and signal intensity. (C) Time course of appearance of recombinant Neor overlap DNA, after transfection of MRC5 (_) and MRC5-SV ( ) cells with DR1 plasmid, monitored by quantitative PCR. The ordinate indicates the recombinant/ control ratio of recovered DNA regions, determined from autoradiographic intensities of PCR-amplified bands at 0.49 and 0.35 kbp, each within the near-linear range of response to DNA input, 1 nM concentration. After DNA amplification, limited to 13 cycles in order to stay within the near-linear range of response to DNA input (see above), samples were analyzed by agarose gel electrophoresis, alkaline transfer to charged nylon membranes (31), and hybridization (13) to a [32P]DNA Neor probe. Recombinant DNA bands at 0.49 kbp were quantitated by densitometry of autoradiographs and comparison with standard curves generated concurrently with pSV2neo. Results were then normalized to the coamplified 0.35-kbp reference (unrecombined) bands similarly quantitated, which served as internal controls. The increase in intensity of the 0.49-kbp band, relative to that of the 0.35-kbp band, yielded time courses for MRC5 and MRC5-SV cells (Fig. 6C) similar to those obtained for plasmid recombination by the bacterial transformation assay (Fig. 2C). These data indicate that the fraction of recombined molecules was elevated 9- to 13-fold in MRC5-SV extracts, relative to that in MRC5, at 36 and 48 h posttransfection (Fig. 6C).
DISCUSSION Immortally transformed cells mediate higher levels of extrachromosomal DNA rearrangement than normal diploid cells. All five immortal human cell lines tested were significantly higher than each of three normal human strains in the bacterial assay of homology-dependent Neor gene rearrangements (Table 1). These comparisons included MRC5-SV, an SV40-transformed cell line which generated 10- to 20-fold more Neor recombinant plasmids than did MRC-5 cells, the normal strain from which it was derived. Clearly, there is a profound difference between these transformed cells and normal diploid cells in their processing of plasmid DNA containing direct repeats. The basis for this difference remains to be determined. We have used the term plasmid recombination to describe such homology-dependent DNA rearrangement leading to Neor plasmids, but this is not intended to imply that homologous recombination is known to be the mechanism involved. We initially considered that the transformed cell lines might differ from normal cells only in the endonucleolytic scission of DNA, which is thought to initiate homologous recombination (25, 45) and enhances the occurrence of both intra- and intermolecular events in mammalian cells and in bacteria (5, 34, 39). Two lines of evidence argue against this. First, there was no correlation evident between the extent of plasmid degradation revealed by Southern blot hybridization
4015
(Fig. 3) and the frequency of plasmid recombination as measured by either assay. Indeed, Werner syndrome fibroblasts, which exhibited the lowest degradation of input plasmid in our experience, were among the most recombinogenic cells assayed (R. Z. Cheng, B. W. Kurz, and R. J. Shmookler Reis, manuscript in preparation). Second, the efficient removal of linear molecules from DNA recovered after passage through either MRC5 or MRC5-SV cells had no effect on the proportion of Neor colonies (Fig. 4B), thus excluding the possibility that DH5 bacteria, although stringently RecA-, complete recombinations initiated in mammalian cells through DNA linearization. Analysis of the putative recombinant region, through a quantitative PCR amplification procedure, confirmed that recombinationlike events in transfected DNA are markedly increased in immortally transformed cells (Fig. 6). The generation of a sharp, discrete band at 0.49 kbp from Hirt supernatant samples (Fig. 6A) argues strongly that amplifiable sequences, when smaller than the 3.7-kbp E. coli gpt-Neor region initially situated between the primers, are predominantly the products of a precise, homology-dependent process such as gene conversion or homologous recombination. We do not exclude a nonconservative mechanism, such as homologous repair of damaged plasmid termini, as a possible pathway for creating recombinant molecules; indeed, the importance of such mechanisms has been demonstrated in several nonhuman immortal cell lines (3, 12, 25, 48). With regard to the 0.49-kbp recombinant region generated in mammalian cells, as assessed by extension of flanking primers (PCR), we must emphasize that at present we are unable to evaluate the relative contributions of homologous recombination, gene conversion, and terminal pairing. Any or all of these processes, of course, could contribute to the karyotypic instability of immortal cells. The cell factors responsible for the observed increase in plasmid recombination in immortally transformed cells remain unknown. Plasmid replication does not appear to occur to a significant degree in the human cell types tested. SV40 T antigen, although altering the interaction of multiple transbinding factors with DNA (15, 38), does not promote plasmid recombination either directly or indirectly, since the observed rate was not increased in cells known to maintain high levels of T antigen, relative to that of the parental T-antigen-lacking cell line. Moreover, a deletion spanning most of the T-antigen-binding domain in DR1 did not reduce Neor region recombination but rather increased it moderately in the presence of T antigen. This downstream inhibition of recombination may result from local DNA unwinding at the T-antigen-binding region (7), which could impede strand separation elsewhere in the molecule by attenuating its negative superhelicity. The kinetics of appearance of recombinant molecules in MRC5-SV cells (Fig. 2C and 6C) indicate a latency period, which may be longer for the production of viable recombinant plasmids than for the local events scored by PCR. This delay, observed in several independent experiments, suggests a possible requirement for entry of cells into S phase before plasmid recombination or resolution can occur. Indeed, a peak of [3H]thymidine incorporation into MRC5-SV cells just preceded the rise in recombination at 24 to 36 h after transfection; however, we noted that recombination was not impaired and in some cases could be elevated, in very slowly cycling cells (data not shown). Implications. A substantial body of evidence indicates that DNA rearrangement may be a frequent and perhaps necessary event in a multistep process leading to oncogenic
4016
FINN ET AL.
transformation (6, 11, 14, 24, 33, 51, 53). Several phenomena mediated by recombinationlike mechanisms (i.e., gene amplification, deletion, and translocation) are prevalent in neoplastic cells (6, 11, 24, 53). Tumor-specific induction of proto-oncogenes has been shown to involve DNA rearrangements which place a strong promoter-enhancer proximal to a proto-oncogene (6, 22), gene amplification events (22, 37), and other uncharacterized chromosomal translocations (6, 53). Moreover, several carcinogens have been demonstrated to have recombinogenic activity (30, 51). It is thus possible that an increase in the rate of homologous recombination or of other DNA-rejoining events might be an early, precrisis step in immortalization by DNA viruses, facilitating subsequent events which are jointly required to confer the stably transformed phenotype. To our knowledge, this is the first direct comparison of recombination frequencies in normal versus immortal cells. Although homologous recombination and gene conversion have been assessed at markers stably integrated into chromosomes of immortal cells (21, 26-28, 30, 35, 39-43, 47, 50), the loci of integration vary from one such subline to another. The principal obstacle to measuring chromosomal recombination in diploid cells, however, is their limited replicative potential, much of which would be exhausted in obtaining suitably engineered clones for assay by existing methodology. Work is in progress to develop new techniques for the rapid determination of chromosomal recombination frequency, which would then obviate the need for plasmid recombination assays to evaluate diploid cell strains. Three of the five human permanent cell lines we tested (MRC5-SV, XPA-SV, and 293) are known DNA virus transformants; moreover, HeLa cells contain integrated papillomavirus 11 DNA, in common with a variety of other cervical carcinomas (8), and may therefore be considered virally transformed. Our data thus indicate an elevation of plasmid recombination associated with transformation to immortality by three DNA tumor viruses from two viral classes (papovaviruses and adenoviruses), as well as one spontaneous tumor cell line of unknown origin (CACL). While this sample is too small to justify general conclusions, the diversity of transforming agents suggest that the changes underlying increased plasmid recombination may reflect a common early event in transformation to immortality. Whether this is a necessary or sufficient condition for transformation by DNA viruses and whether it extends also to retrovirally or chemically induced transformants are questions requiring further study. In any case, such data can only suggest an association between immortal transformation of cells and a recombinationlike process and indicate the need for more incisive experiments to test whether recombinogenic factors play a causal role in cell transformation or in the etiology of cancer. ACKNOWLEDGMENTS We thank Elena Moerman for expert technical assistance with tissue culture, Hans Joenje for HeLa cells, Suresh Subramani for the DR1 plasmid, and Raju Kucherlapati, Suresh Subramani, and David Gelfand for helpful discussions. LITERATURE CITED 1. Allard, D., L. Delbecchi, D. Bourgaux-Ramoisy, and P. Bourgaux. 1988. Major rearrangement of cellular DNA in the vicinity
of integrated polyomavirus DNA. Virology 162:128-136.
2. Anai, M., T. Hirahashi, and Y. Takagi. 1970. A deoxyribonuclease which requires nucleoside triphosphate from Micrococcus lysodeikticus. J. Biol. Chem. 245:767-774. 3. Anderson, R. A., and S. L. Eliason. 1986. Recombination of
MOL. CELL. BIOL. homologous DNA fragments transfected into mammalian cells occurs predominantly by terminal pairing. Mol. Cell. Biol. 6:3246-3252. 4. Ayares, D., L. Cherkuri, K.-Y. Song, and R. Kucherlapati. 1986. Sequence homology requirements for intermolecular recombination in mammalian cells. Proc. Natl. Acad. Sci. USA 83: 5199-5203. 5. Ayares, D., J. Spencer, F. Schwartz, B. Morse, and R. Kucherlapati. 1985. Homologous recombination between autonomously replicating plasmids in mammalian cells. Genetics 111:375-388. 6. Bishop, J. M. 1987. The molecular genetics of cancer. Science 235:305-311. 7. Borowiec, J. A., and J. Hurwitz. 1988. Localized melting and structural changes in the SV40 origin of replication induced by T-antigen. EMBO J. 7:3149-3158. 8. Boshart, M., L. Gissman, H. Ikenberg, A. Kleinheinz, W. Scheurlen, and H. zur Hausen. 1984. A new type of papillomavirus DNA, its presence in genital cancer biopsies and in cell lines derived from cervical cancer. EMBO J. 3:1151-1157. 9. Bradley, W. E. C., J. L. P. Gareau, A. M. Seifert, and K. Messing. 1987. Molecular characterization of 15 rearrangements among 90 human in vivo somatic mutants shows that deletions predominate. Mol. Cell. Biol. 7:956-960. 10. Butner, K. A., and C. W. Lo. 1986. High frequency DNA rearrangements associated with mouse centromeric satellite DNA. J. Mol. Biol. 187:547-556. 11. Cairns, J. 1981. The origin of human cancers. Nature (London) 289:353-357. 12. Chakrabarti, S., and M. M. Seidman. 1986. Intramolecular recombination between transfected repeated sequences in mammalian cells is nonconservative. Mol. Cell. Biol. 6:2520-2526. 13. Church, G. M., and W. Gilbert. 1984. Genomic sequencing. Proc. Natl. Acad. Sci. USA 81:1991-1995. 14. Fidler, I. J. 1985. Genetic mechanisms in tumor progression, heterogeneity, and metastasis, p. 221-231. In B. Pullman, P. 0. Ts'o, and E. L. Schneider (ed.), Interrelationship among aging, cancer and differentiation. D. Reidel Publishing Co., Dordrecht, The Netherlands. 15. Gallo, G. J., G. Gilinger, and J. C. Alwine. 1988. Simian virus 40 T antigen alters the binding characteristics of specific simian DNA-binding factors. Mol. Cell. Biol. 8:1648-1656. 16. Gluzman, Y., J. F. Sambrook, and R. J. Frisque. 1980. Expression of early genes of origin-defective mutants of SV40. Proc. Natl. Acad. Sci. USA 77:3898-3902. 17. Goldberg, I., and J. J. Mekalanos. 1986. Effect of a recA mutation on cholera toxin gene amplification and deletion events. J. Bacteriol. 165:723-731. 18. Hanahan, D. 1983. Studies on transfection of Escherichia coli with plasmids. J. Mol. Biol. 166:557-580. 19. Hirt, B. 1967. Selective extraction of polyoma DNA from infected mouse cell cultures. J. Mol. Biol. 26:365-369. 20. Hoel, P. G. 1962. Introduction to mathematical statistics, p. 279. John Wiley & Sons, Inc., New York. 21. Jasin, M., J. DeVilliers, F. Weber, and W. Schaffner. 1985. High frequency of homologous recombination in mammalian cells between endogenous and introduced SV40 genomes. Cell 43: 696-703. 22. Land, H., L. F. Parada, and R. A. Weinberg. 1983. Cellular oncogenes and multistep carcinogenesis. Science 222:771-778. 23. Lave, L. B., and G. S. Omenn. 1986. Cost-effectiveness of short-term tests for carcinogenicity. Nature (London) 324:2934. 24. Leder, P., J. Battey, and G. Lenoir. 1983. Translocations among antibody genes in human cancer. Science 222:765-771. 25. Lin, F. W., K. Sperle, and N. Sternberg. 1984. Model for homologous recombination during transfer of DNA into mouse L cells: role for DNA ends in the recombination process. Mol. Cell. Biol. 4:1020-1034. 26. Lin, F. W., and N. Sternberg. 1984. Homologous recombination between overlapping thymidine kinase gene fragments stably inserted into a mouse cell genome. Mol. Cell. Biol. 4:852-860. 27. Liskay, R. M., A. Letsou, and J. L. Stachelek. 1987. Homology requirement for efficient gene conversion between duplicated
VOL. 9. 1989
28.
29.
30.
31. 32.
33. 34.
35. 36.
37.
38. 39.
40.
PLASMID RECOMBINATION IN NORMAL AND TRANSFORMED CELLS
chromosomal sequences in mammalian cells. Genetics 115: 161-167. Liskay, R. M., and J. L. Stachelek. 1986. Information transfer between duplicated chromosomal sequences in mammalian cells involves contiguous regions of DNA. Proc. NatI. Acad. Sci. USA 83:1802-1806. Lopata, M. A., D. W. Cleveland, and B. Sollner-Webb. 1984. High level transient expression of chloramphenicol acetyl transferase gene by DEAE-dextran mediated DNA transfected coupled with a dimethyl sulfoxide or glycerol shock treatment. Nucleic Acids Res. 12:5707-5717. Luisi-Deluca, C., R. D. Porter, and W. D. Taylor. 1984. Stimulation of recombination between homologous sequences on plasmid DNA and chromosomal DNA in Escherichia coli by N-acetoxy-2-acetyl-aminofluorene. Proc. Natl. Acad. Sci. USA 81:2831-2835. Reed, K. C., and D. A. Mann. 1985. Rapid transfer of DNA from agarose gels to nylon membranes. Nucleic Acids Res. 13: 7207-7221. Rosen, S. W., B. D. Weintraub, and S. A. Aaronson. 1980. Nonrandom ectopic protein production by malignant cells: direct evidence in vitro. J. Clin. Endocrinol. Metab. 50:834-841. Rubin, H. 1985. Cancer as a dynamic developmental disorder. Cancer Res. 45:2935-2942. Rubnitz, J., and S. Subramani. 1985. Rapid assay for extrachromosomal homologous recombination in monkey cells. Mol. Cell. Biol. 5:529-537. Rubnitz, J., and S. Subramani. 1987. Correction of deletions in mammalian cells by gene conversion. Somatic Cell Mol. Genet. 13:183-190. Saiki, R. K., D. Gelfand, S. Stoffel, S. J. Scharf, R. Higuchi, G. T. Horn, K. Mullis, and H. A. Erlich. 1988. Primer-directed enzymatic amplification of DNA with a thermostable DNA polymerase. Science 239:487-494. Slamon, D. J., G. M. Clark, S. G. Wong, W. J. Levin, A. Ulrich, and W. L. McGuire. 1987. Human breast cancer: correlation of relapse and survival with amplification of the HER-2/neu oncogene. Science 235:177-182. Smale, S. T., and R. Tjian. 1986. T-antigen-DNA polymerase at complex implicated in simian virus 40 DNA replication. Mol. Cell. Biol. 6:4077-4087. Smithies, O., R. Gregg, S. Boggs, M. Koralewski, and R. S. Kucherlapati. 1985. Insertion of DNA sequences into the human chromosomal gamma-globin locus by homologous recombination. Nature (London) 317:230-234. Stahl, F. W. 1986. Roles of double-strand breaks in generalized
41.
42.
43. 44. 45.
46.
47. 48. 49.
50.
51.
52.
53.
4017
genetic recombination. Prog. Nucleic Acid Res. Mol. Biol. 33:169-193. Stringer, J., R. Kuhn, J. Newman, and J. Meade. 1985. Unequal homologous recombination between tandemly arranged sequences stably incorporated into cultured rat cells. Mol. Cell. Biol. 5:2613-2622. Subramani, S., and P. Berg. 1983. Homologous and nonhomologous recombination in monkey cells. Mol. Cell. Biol. 3: 1040-1052. Subramani, S., and J. Rubnitz. 1985. Recombination events after transient infection and stable integration of DNA into mouse cells. Mol. Cell. Biol. 5:659-666. Sussman, D., and G. Milman. 1984. Short-term, high-efficiency expression of transfected DNA. Mol. Cell. Biol. 4:1641-1643. Szostack, J. W., T. L. Orr-Weaver, R. J. Rothstein, and F. W. Stahl. 1983. The double-strand break repair model for recombination. Cell 33:25-35. Tennant, R. W., B. H. Margolin, M. D. Shelby, E. Zeiger, J. K. Haseman, J. Spalding, W. Caspary, M. Resnick, S. Stasiewicz, B. Anderson, and R. Minor. 1987. Prediction of chemical carcinogenicity in rodents from in vitro genetic toxicity assays. Science 236:933-940. Thomas, K. R., K. R. Folger, and M. R. Capecchi. 1986. High frequency targeting of genes to specific sites in the mammalian genome. Cell 44:419-428. Wake, C. T., T. Gudewiez, T. Porter, A. White, and J. H. Wilson. 1984. How damaged is the biologically active subpopulation of transfected DNA? Mol. Cell. Biol. 4:387-398. Wake, C., F. Vernaleone, and J. H. Wilson. 1985. Topological requirements for homologous recombination among DNA molecules transfected into mammalian cells. Mol. Cell. Biol. 5: 2080-2089. Waldman, A. S., and R. M. Liskay. 1987. Differential effects of base-pair mismatch on intrachromosomal versus extrachromosomal recombination in mouse cells. Proc. Natl. Acad. Sci. USA 84:5340-5344. Wang, Y., V. M. Maher, R. M. Liskay, and J. J. McCormick. 1988. Carcinogens can induce homologous recombination between duplicated chromosomal sequences in mouse L cells. Mol. Cell. Biol. 8:196-202. Yamagishi, H., T. Tsuda, S. Fujimoto, M. Toda, K. Kato, Y. Maekawa, M. Umeno, and M. Anai. 1983. Purification of polydisperse circular DNA of eukaryotic cells by use of ATPdependent deoxyribonuclease. Gene 26:317-321. Yunis, J. J. 1987. Multiple recurrent genomic rearrangements and fragile sites in human cancer. Somatic Cell Mol. Genet. 13:397-403.
