Although Gonda and Radding (13) could not observe in vitro RecA-promoted pairing of molecules sharing only 30 bp of homologous sequence, recent in vivo.
Proc. Nati. Acad. Sci. USA Vol. 83, pp. 5199-5203, July 1986 Genetics
Sequence homology requirements for intermolecular recombination in mammalian cells (genetic recombination/human cells/gene conversion)
DAVID AYARES, LAVANYA CHEKURI, KYU-YOUNG SONG, AND RAJU KUCHERLAPATI* Center for Genetics, University of Illinois at Chicago, College of Medicine, Chicago, IL 60612
Communicated by J. T. Bonner, March 24, 1986
We have examined the homology requireABSTRACT ments for intermolecular recombination between plasmids introduced into human, monkey, and bacterial cells. Variablesize-deletion derivatives of the prokaryotic-eukaryotic shuttle vector pSV2neo were constructed. Each of these plasmids was mixed with another pSV2neo plasmid containing a different, nonoverlapping deletion. Recombination was measured in mammalian cells and bacteria by the frequency of reconstruction of an intact neo gene. We observed that 25 base pairs of homologous sequence is sufficient -to yield recombinant products, implying that synapsis and homologous pairing can occur with this level of homology. Examination of the products revealed that nonreciprocal recombination played a role in the generation of normal neo genes. In addition coconversion of linked markers was observed. Exonucleoiytic action seems to play a role in gene conversion. It has been shown recently that somatic mammalian cells possess the enzymatic machinery required to mediate general recombination between homologous sequences (1-9). Though the mechanisms of this process have yet to be elucidated, it must require exchange of genetic information between homologous DNA sequences. Such a homologydependent synapsis reaction is an essential component of all proposed models for homologous recombination (10-12). Indeed, bacterial RecA protein, a key enzyme in recombination, is capable of mediating strand exchange between homologous sequences (13). Sequence homology is not only necessary for the production of a nascent heteroduplex during homologous pairing but also for the process of branch migration because joint sites cannot traverse nonhomologous regions of DNA (14, 15). Homologous sequences have also been shown to be important for site-specific recombination in a number of systems in bacteria and fungi (16-21). The obvious requirement of sequence homology for recombination in prokaryotes and eukaryotes led to investigations of the degree of sequence homology required for synapsis during intermolecular recombination. In bacteria, Singer et al. (22) found that a minimum of 50 base pairs (bp) of homologous sequence was required for recombination between two mutants with deletions in the r1I cistrons of bacteriophage T4. Although Gonda and Radding (13) could not observe in vitro RecA-promoted pairing of molecules sharing only 30 bp of homologous sequence, recent in vivo experiments using phage vectors revealed that as few as 20 bp, but not 16 bp, of homologous sequence resulted in recombination (23). Rubnitz and Subramani (24) investigated the homology requirements for intramolecular recombination between homologous simian virus 40 (SV40) sequences in monkey CV-1 The publication costs of this article were defrayed in part by page charge payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact.
5199
cells. By progressively decreasing the length of homologous sequence between the SV40 sequences on the vector and using a plaque assay to measure the frequency with which infectious virions were generated, they have shown that low levels of recombination could be detected when there was only 14 bp of homologous sequence. Though reports ofhomology requirements for intermolecular recombination in mammalian cells have appeared (1, 25), no systematic investigations have been conducted. Since the mechanisms of intramolecular and intermolecular recombination events may be different (26), we have studied the minimum amount of sequence homology required for intermolecular recombination in mammalian cells. We have generated a series of mutants that contain variable size deletions of the neomycin phosphotransferase gene (neo) in the eukaryotic-prokaryotic shuttle vector pSV2neo (27). Using each of these plasmids as a partner in homologous recombination in different mammalian cells, we now show that as little as 25 bp of homologous sequence is adequate for homologous recombination. Examination of the products of recombination has provided information about some of the steps involved in the generation of recombinant molecules.
MATERIALS AND METHODS Cells. EJ human bladder carcinoma cells and COS-1 SV40-transformed monkey kidney cells (28) were grown and maintained in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. Selection for neo expression was achieved by plating cells in medium containing the aminoglycoside analog G418 (GIBCO) at 400 ,g/ml. The RecAEscherichia coli strain DH1 was used for all bacterial transformations. A RecA+ strain of E. coli, RR1, was used to study homology requirements in bacteria. Plasmid Substrates for Recombination. The substrates for recombination are deletion mutants of the plasmid pSV2neo (27), shown in Fig. 1. pSV2neo contains the ampicillin-resistance gene, a bacterial replication origin from pBR322 and the SV40 replication origin. In addition, it also contains the neo gene from Tn5, which confers resistance to kanamycin or neomycin in bacteria and resistance to G418 in mammalian cells. This plasmid replicates autonomously in bacterial and monkey COS cells and integrates into the chromosomal DNA in EJ cells. pLCK101 is derived from pSV2neo-DL, a plasmid containing a 248-bp deletion at the 5' end of the coding region of the neo gene in pSV2neo. pSV2neo-DR-SSaLX contains a 283-bp deletion at the 3' end of the neo gene. The construction of these deletion plasmids was described by Kucherlapati et al. (6) and Song et al. (29). Deletions resulting in decreasing Abbreviations: bp, base pair(s); SV40, simian virus 40; NeoR, neomycin resistant; AmpR, ampicillin resistant. *To whom correspondence should be addressed at: Center for Genetics, University of Illinois at Chicago, College of Medicine, 808 South Wood Street, Chicago, IL 60612.
5200
Genetics: Ayares
et
al.
Proc. Natl. Acad. Sci. USA 83 (1986) K
E
B
B
pLCK1O1
DR SSaIX S Bg E
S
I Sal
S Bg E
E
I
H
Xba
BgSa
S
I Digest B
Xba Bal 31 exonuclease
E
E
...n
S Bg
x
(12)
x
FIG. 1. Substrates used for recombination. The construction of the plasmid substrates pSV2neo-DR-SSalX and pLCK101 has been described (29). Derivatives of pSV2neo-DR-SSa1X were obtained as described in the text. A total of 12 DR der plasmids were used in the experiments. B, BamHI; Bg, Bgl II; E, EcoRI; H, HindIII; K, Kpn I; Sal, Sal I; S, Sma I; Xba, Xba I; X, Xho I. amounts of homology available for homologous recombination
generated in the following fashion. pSV2neo-DR-SSaLX DNA was digested at the site of the original deletion with Sal I to completion and treated with BAL-31 nuclease for 5-9 min; the products were pooled. Xho I linkers were added at the ends generated by BAL-31 digestion, and the plasmids were religated and used to transform E. coli strain HB101 by methods of Maniatis et al. (30). Plasrnid DNA was prepared by the alkaline lysis method (30). The plasmid DNA was analyzed by digestion with restriction endonucleases Pst I or Bgl II/Xho I to determine the extent of the deletion. The plasmids are designated as DR der plasmids. Transfections and Assay for Recombination. EJ cells. Two micrograms of pSV2neo-DL-derived pLCK101 linearized with Sal I was mixed with 2 jig each of the DR der plasmids in the absence of carrier DNA and cotransfected by the calcium phosphate precipitation method (31). Transfected cells were plated in medium containing G418, and colonies were scored 14 days after transfection. COS cells. Eight micrograms of each of the two deletion plasmids were mixed and cotransfected into COS-1 cells by the DEAE-dextran-mediated DNA transfer method (32, 33). Low molecular weight DNA was isolated 48 hr after transfection by the method of Hirt (34) and was used to transform a RecA- strain of E. coli (DH1), which was plated on ampicillin- or neomycin-containing plates. Recombination frequency is expressed as the total number of neomycin., resistant (NeoR) colonies over the number of ampicillinresistant (AmpR) colonies. To measure recombination in bacteria, mixtures of appropriate pairs of substrates were presented to the RRI strain of E. coli. Recombination frequency is expressed as above. were
RESULTS Generation of Deletion Substrates. We have previously shown efficient recombination between pSV2neo-DL and pSV2neo-DR in several different mammalian cell types (6, 8). The end points of the deletions in these plasmids are 501 bp apart. Thus, the degree of homology within which recombi-
nation has to occur to give rise to a wild-type neo gene is 501 bp. We generated a series of derivative plasmids from pSV2neo-DR-SSalX as described. From a large series of plasmids, designated DR der plasmids, we chose 12 for detailed study. The extent of homology that each of these 12 plasmids had with pLCK101 was determined by digestion of each of the DR der plasmids with different restriction endonucleases and fractionating the products on polyacrylamide gels. Results of this experiment are shown in Fig. 2. The extent of the deletions were such that the sequence homology ranged from 501 bp to 25 bp. In each case, the standard error is 5 bp. Determination of Sequence Homology Requirements in Mammallan Cells. In order to examine the degree of sequence homology required for the recombination in mammalian cells, substrates with decreasing amounts of homologous sequence were used to transfect either EJ human bladder carcinoma cells or monkey COS-1 cells. We used Sal I-linearized pLCK101 as one of the substrates because such linearization increases recombination (8). Sal I cuts at the site of the deletion. This plasmid was mixed with the appropriate DR der plasmid and transfected into human EJ cells by calcium phosphate coprecipitation or into COS cells (using DEAE-dextran). Recombination frequency was determined as described. Results obtained from transfection into the different cell types are summarized in Table 1, and the relative recombination frequencies are shown in Fig. 3. In EJ cells, mixtures of DL and DR SSaIX plasmids yielded 16.4 colonies per ug of input DNA. When the homologous sequence was reduced to 415 bp, the yield of recombinants was decreased by a factor of 4. The level of recombination sharply declined when the homologous sequence was reduced to 330 bp (2.4% of that observed with 500 bp of homologous sequence). At levels of homologous sequence less than 330 bp, we were able to detect a low but significant level of recombination. Homologous sequence as low as 25 bp was found to be sufficient for recombination. The relationship of recombination to the degree of sequence homology was biphasic in nature (Fig. 3). The results obtained in experiments with the COS cells were somewhat different than those observed with the human EJ ±
Genetics: Ayares
et
al.
Proc. Natl. Acad. Sci. USA 83 (1986)
5201
1
100
80
0
Ii ~60 E 0
>40
0
0
0
co ~0
0-
20~~~
0
FIG. 2. Analysis of the DR der plasmids. DR der plasmids were constructed as described in the text. Each of the plasmid DNAs was digested to completion with Bgl II and Xho I and fractionated on 12% polyacrylamide gels. The extent of homology in the neo gene sequence between pLCK101 and each of the DR der plasmids was obtained by deducting 165 bp from the size of the smaller gel fragment; 165 bp is the distance from the Bgl II site to the site corresponding to the Sal I site in pLCK101. Lanes: A, pBR322 DNA digested with Hpa II; B, pBR322 digested with Hae III; C-O, plasmid DNA digested with Bgl II/Xho I; C, pSV2neo-DR-SSalX; D, DR der 415; E, DR der 330; F, DR der 315; G, DR der 265; H, DR der 227; I, DR der 205; J, DR der 190; K, DR der 140; L, DR der 125; M, DR der 110; N, DR der 35; and 0, DR der 25. Numbers on the right are lengths in bp, based on the marker fragments in lanes A and B.
cells (Table 1 and Fig. 3). As in the case of El cells, COS cells able to mediate recombination between the exogenously introduced plasmids. As a control, each of the pair of recombination substrates was transfected separately into COS cells, and the resulting low molecular weight DNA was mixed prior to transformation of E. coli. No NeoR colonies were obtained from this experiment (data not shown). This result indicated that the NeoR colonies were the result ofrecombination in COS cells and not in the recombination-deficient bacteria. The recombination frequency in COS cells ranged from 1.79 x 10-2 for 501 bp of homologous sequence to 2.0 x 1i-4 for 25 bp of homologous sequence. When the relative rates of recombination were plotted (recombination with 500 bp of homologous sequence was considered 100%), we observed that there is a linear relationship between the extent of homology and frewere
10;0
0
200
300
Homologous
sequence,
400
500
bp
Relative rates of recombination in EJ, COS, and bacterial in Table 1 shown used to generate the this figure. Recombination obtained with 500 bp of homologous
FIG. 3.
cells. in
Data
sequence
shown
was
curves
were
considered 100%. o, COS cells; e, EJ cells; andA, E.
coli.
quency
of
recombination. This
result
contrasts
with that ob-
tained from EJ cells.
homology requirements for recombination in bacteria by Watt et al. (23), and they showed that 20 bp of homologous sequence was adequate for homologous recombination whereas 16 bp was not sufficient. The
have been studied in detail
Because the plasmid substrates we used contained deletions in the neo gene, which can be selected for in bacteria as well as in mammalian cells, we were able to compare the efficiency of recombination in these two systems. Pairs of appropriate plasmids were transfected into a RecA' strain of E. coli, RRI. The transformation mixture was plated on plates containing neomycin and ampicillin, and the recombination frequency was determined as before. The frequency of recombination in these bacteria was at least 1/100th that observed in the mammalian COS cells (see Table 1). Recombination in bacteria ranged from 1 x 10-4 at 501 bp of homologous sequence to 9 x 10-6 at 25 bp of homologous sequence. These results are in agreement with those obtained by others (22, 23) and indicate that 25 bp are sufficient to
Table 1. Relationship between the extent of sequence homology and recombination frequency in mammalian and bacterial cells Bacteria COS cells EJ cells NeoR NeoR Homologous NeoR/AmpR NeoR/AmpR Colonies/,ug x 10-5 x 10-3 sequence, bp Exp., no. colonies Exp., no. colonies of DNA Exp., no. Frequency 501 2 4 194 11.2 862 16.4 17.9 9 458/24 415 2 85 6.7 4 735 14.8 4.2 6 50/12 330 2 52 5.1 4 0.4 667 14.3 4 6/14 315 2 48 4 3.8 464 7.2 5 0.3 5/18 265 2 52 4 6.3 368 7.1 0.8 6 18/22 2 227 40 3.1 4 270 7.4 2.6 6 42/16 205 2 29 2.3 2 240 0.6 7.1 4 8/14 2 45 190 4 3.5 168 5.2 7 1.3 31/24 140 2 28 2 2.4 105 0.5 4.9 1 2/4 125 2 21 4 118 0.4 1.8 2.3 8 11/26 110 2 27 4 2.2 87 0.8 2.7 7 17/24 2 35 12 2 1.0 26 1.4 5 0.2 3/14 25 2 11 2 8 8 0.3 0.9 0.2 8/26
5202
Genetics:
Ayares et al.
Proc. Natl. Acad. Sci. USA 83 (1986) Table 2. Relationship between the degree of sequence homology and coconversion of linked markers
allow recombination and, as such, homologous alignment and pairing in both bacteria and mammalian cells. Products of Recombination. We have deduced valuable information about the mechanism of homologous recombination by analyzing the plasmid DNA harbored by NeoR colonies obtained from the COS cell experiments (8). In the current experiments, the fact that there are several restriction enzyme site polymorphisms in the two substrates proved useful in deducing the nature of the recombination event. pLCK101 contained a HindIII site at the 5' end of the neo gene and a Sma I site at the 3' end. pSV2neo-DR-SSalX contained a Sma I site at the 5' end and a Xba I site at the 3' end (Fig. 1). In the DR der plasmids, the Sma I site was unaltered while the Xba I site was lost in all except the pSV2neo der 415 plasmid. Examination of the plasmids we have obtained revealed them to be either dimers or monomers. The dimeric molecules contained a wild-type neo gene and a deleted neo gene that corresponded in size to that present in the DR der plasmid. These results indicate that, during recombination, the deleted neo gene in pLCK101 was converted to wild type, followed by a crossing-over event. Similar results were obtained when DL and DR plasmids were used as substrates (8). Since gene conversion was playing an important role in recombination, we examined whether the extent of gene conversion is in some way related to the degree of sequence homology. We have previously noted that conversion of the deletion in DL plasmids is accompanied by the coconversion of the adjacent 5' HindIII site to a Sma I site (8). Such a coconversion event can be detected readily by measuring the number of Sma I sites in monomeric recombination products. Presence of a single Sma I site indicates no coconversion, while the presence of two Sma I sites is indicative of coconversion. Results obtained from such an analysis are summarized in Table 2. The predominant class of monomeric plasmids, obtained with plasmids having homologous sequences down to 110 bp, contained two Sma I sites (i.e., coconversion of the 5' HindIII site). The situation changed dramatically when the homologous sequence was reduced to 35 or 25 bp. The majority of the monomeric plasmids from these crosses contained a single Sma I site at the 3' end of the neo gene, while the 5' HindIII site was unchanged. The possible significance of these results to an understanding of the mechanism of recombination in mammalian cells is discussed later in this report.
Homologous Plasmids with Sma I sites, no. 2 sites 1 site sequence, bp Total 501 23 2 25 140 11 1 12 125 9 3 12 12 110 12 0 35 0 12 12 25 1 9 10 Products of recombination obtained from experiments with plasmids with different amounts of homologous sequence. Plasmid DNA was isolated from NeoR colonies after transformation of the RecAE. coli strain DH-1 with low molecular weight DNA from cotransfection of COS-1 cells.
DISCUSSION We have examined the degree of sequence homology required for pairing of complementary DNAs during intermolecular recombination in human and monkey cells as well as in bacteria. Pairs of plasmids bearing different degrees of homology were introduced into mammalian cells, and recombination was monitored. In both human EJ cells and monkey COS cells, there was a decrease in the frequency of recombination with a decreasing length of homologous sequence, indicating that the mechanism of synapsis of complementary sequences in mammalian cell recombination is a homologydependent process. However, the pattern of decrease in recombination observed with diminishing sequence homology was different in EJ and COS cells (Fig. 2). In COS cells, there was a linear relationship between recombination frequency and sequence homology, whereas in human EJ cells recombination frequency dropped at a faster rate and the shape of the curve showed evidence of a biphasic relationship. An important difference between the two systems is that the input plasmids can replicate autonomously in COS cells and cannot do so in EJ cells. Thus, it is possible that the linear relationship between the length of homologous sequence and the recombination frequency is somehow related to the continuous replication of the plasmids. We have not been able to establish the nature of this relationship. By using the same deletion substrates, the sequence homology requirement for recombination in RecA' E. coli was also studied. The frequency of recombination obtained for recom-
A H
S
B S
1110bpl\
X
Exonuclease
S
H s
bp
\I
x
/
Exonuclease
No exonuclease
S
S
S
Recombination
S
Si,,
x
No recombination
S
I
Recombination H
S
FIG. 4. A model to explain the relationship between sequence homology and gene conversion. (A) When there is extensive sequence homology, the degree of exonucleolytic digestion that results in the loss of the 5' flanking marker does not result in loss of homologous sequences needed for recombination and generation of a wild-type neo gene. (B) When the overall level of homology is reduced, exonucleolytic digestion resulting in loss of 5' flanking marker results in complete loss of sequence homology to generate a wild-type neo gene.
Genetics: Ayares
et
al.
bination in bacteria was approximately 1/100th that observed in mammalian cells for all lengths of homologous sequence (Table 1) and was in agreement with previous studies of recombination in prokaryotes (22, 23). The low level of recombination may be attributed to a low frequency of cotransformation of bacterial cells. Although there was a great difference in the frequencies of homologous recombination in eukaryotes versus prokaryotes, the sequence homology requirements for the recombination were the same. A low but significant number of NeoR colonies were obtained in both the RRI RecA' bacteria and in the two mammalian cell types when as little as 25 bp of homologous sequence was shared by the two deletion substrates. Bacterial cells and human cells showed a biphasic relationship between the degree of homology and recombination (Fig. 3). Rubnitz and Subramani investigated the sequence homology requirements for intramolecular recombination in monkey CV-1 cells and showed that 14 bp of homologous sequence could mediate the pairing of homologous sequences on the same plasmid molecule (24). The substrates used in our experiments in monkey COS-1 cells and human EJ cells contained a minimum of 25 bp of homologous sequence, and this amount was sufficient in all the systems tested. Rubnitz and Subramani (24) observed a steep drop in recombination frequency when the homologous sequence decreased from 214 to 163 bp, whereas we find in EJ cells and bacterial cells that the severest break occurred between 501 and 330 bp of homologous sequence. The small differences between the two systems may reflect differences in the substrates used for recombination. Analysis of restriction enzyme sites in recombinant products obtained from experiments with different degrees of sequence homology revealed an interesting feature. When the homologous sequence in the neo gene region was 110 bp or greater, a high proportion of the monomeric plasmids (80-95%) resulted from coconversion of the DL deletion and the 5' HindIII site. When the homologous sequence was 35 bp or lower, wild-type plasmids resulted from repair of the DL deletion, but coconversion of the 5' HindIII site was observed in only 1 of 22 cases. We offer the following explanation for these results. Coconversion of two closely linked markers could result from heteroduplex formation followed by mismatch correction (11) or from exonucleolytic digestion followed by gap repair (12). The coconversion that we observe seems to be the result of exonucleolytic digestion followed by gap repair (Fig. 4). In pLCK101 the HindIII site is 240 bp away from the 5' end point of the DL deletion. When the homology between the two substrate sequences is reduced to 35 bp or lower, an exonucleolytic digestion of 240 bp on either side of the DL deletion would result in a complete loss of homology. When the homology is low, such extensive exonucleolytic digestions would not result in a recombination event leading to a wild-type neo gene. As the homology increases, a greater degree of exonucleolytic digestion can be tolerated, resulting in opportunities for generation of wildtype neo gene along with coconversion of markers close to the deletion. It has to be noted, however, that the substrate that underwent the gene conversion is a "linear gapped duplex" and that the mechanism by which coconversion of genetic markers occurs in this system may not be similar to what happens in meiosis. It is interesting that the sequence homology requirements for recombination in bacteria as well as mammalian cells are very similar. Thomas (35) pointed out that a stretch of 20 bp of homologous sequence is required to form a stable DNA duplex. The requirement of 14 bp of homologous sequence in mammalian cells (24) and of 20 bp in bacteria (23) may reflect this fact. Mammalian genomes contain highly reiterated sequences that are distributed throughout the genome. Because the length of homologous sequence required for recombination is much less than the length of the repeat sequences, it is possible that
Proc. Natl. Acad. Sci. USA 83 (1986)
5203
mammalian systems have developed additional safeguards to prevent recombination between these repeated sequences. Further understanding ofthe mechanism of recombination may reveal the nature of these safeguards. We appreciate many helpful discussions with 0. Smithies and P. Moore. S. Ehrlich provided technical support, and the manuscript was prepared by K. Garwood. This work is supported by grants from the March of Dimes Birth Defects Foundation (1-806) and the National Institutes of Health (GM 33943). 1. Brenner, D. A., Smigocki, A. C. & Camerini-Otero, R. D. (1985) Mol. Cell. Biol. 5, 684-691. 2. DeSaint Vincent, B. R. & Wahl, G. M. (1983) Proc. Natl. Acad. Sci. USA 80, 2002-2006. 3. Folger, K. R., Wong, E. A., Wahl, G. & Capecchi, M. R. (1982) Mol. Cell. Biol. 2, 1372-1387. 4. Wake, C. T. & Wilson, J. H. (1979) Proc. Natl. Acad. Sci. USA 76, 2876-2880. 5. Wasmuth, J. J. & Vock Hall, L. (1984) Cell 36, 697-707. 6. Kucherlapati, R. S., Eves, E. M., Song, K. Y., Morse, B. S. & Smithies, 0. (1984) Proc. Natd. Acad. Sci. USA 81, 3153-3157. 7. Kucherlapati, R. S., Ayares, D., Hannekean, A., Noonan, K., Rauth, S., Spencer, J. M., Wallace, L. & Moore, P. D. (1984) Cold Spring Harbor Symp. Quant. Biol. 49, 191-197. 8. Ayares, D., Spencer, J. M., Schwartz, F., Morse, B. & Kucherlapati, R. S. (1985) Genetics 111, 375-388. 9. Small, J. & Scangos, G. (1983) Science 219, 174-176. 10. Holliday, R. (1964) Genet. Res. 5, 282-304. 11. Meselson, M. S. & Radding, C. M. (1975) Proc. Natl. Acad. Sci. USA 72, 358-361. 12. Szostak, J. W., Orr-Weaver, T. L., Rothstein, R. J. & Stahl, F. (1983) Cell 33, 25-35. 13. Gonda, D. K. & Radding, C. M. (1983) Cell 34, 647-654. 14. Cox, M. M. & Lehman, I. R. (1981) Proc. Natl. Acad. Sci. USA 78, 6018-6022. 15. DasGupta, C. & Radding, C. M. (1982) Proc. Natl. Acad. Sci. USA 79, 762-766. 16. Andrews, B. J., Proteau, G. A., Beatty, L. G. & Sadowski, P. D. (1985) Cell 40, 795-803. 17. Johnson, R. C. & Simon, M. I. (1985) Cell 41, 781-791. 18. Zieg, J. & Simon, M. (1980) Proc. Natl. Acad. Sci. USA 77, 4196-4200. 19. Mizuuchi, K., Weisberg, R., Enquist, L., Mizuuchi, M., Buraczynska, M., Foeller, C., Hsu, P. C., Ross, W. & Landy, A. (1980) Cold Spring Harbor Symp. Quant. Biol. 44, 429-437. 20. Cox, M. M. (1983) Proc. Natl. Acad. Sci. USA 80, 4223-4227. 21. Landy, A. & Ross, W. (1977) Science 197, 1147-1160. 22. Singer, B. S., Gold, L., Gauss, P. & Doherty, D. H. (1982) Cell 31, 25-33. 23. Watt, V. M., Ingles, C. J., Urdea, M. S. & Rutter, W. J. (1985) Proc. Nadl. Acad. Sci. USA 82, 4768-4772. 24. Rubnitz, J. & Subramani, S. (1984) Mol. Cell. Biol. 4, 2253-2258. 25. Miller, C. K. & Temin, H. M. (1983) Science 220, 606-609. 26. Lin, F. W., Sperle, K. & Sternberg, N. (1984) Mol. Cell. Biol. 4, 1020-1034. 27. Southern, P. J. & Berg, P. (1982) J. Mol. Appl. Genet. 1, 327-341. 28. Gluzman, Y. (1982) Cell 23, 175-182. 29. Song, K. Y., Chekuri, L., Rauth, S., Ehrlich, S. & Kucherlapati, R. S. (1985) Mol. Cell. Biol. 5, 3331-3336. 30. Maniatis, T., Fritsch, E. F. & Sambrook, J. (1982) Molecular Cloning: A Laboratory Manual (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY). 31. Graham, F. L. & van der Eb, A. J. (1983) Virology 52, 456-467. 32. Lopata, M. A., Cleveland, D. W. & Sollner-Webb, B. (1984) Nucleic Acids Res. 12, 5707-5717. 33. Sussman, D. J. & Milman, G. (1984) Mol. Cell. Biol. 4, 1641-1643. 34. Hirt, B. (1967) J. Mol. BiD;. 36, 365-369. 35. Thomas, C. A. (1966) Prog. Nucleic Acid Res. Mol. Biol. 5, 315-348.
