Proc. Natl. Acad. Sci. USAVol. 76, No. 9, pp. 4289-4293, September 1979Biochemistry

Mechanism for circularization of linear DNAs: Circular parvovirusMVM DNA is formed by a "noose" sliding in a "lasso"-likeDNA structure

(circle formation/unwinding protein/DNA replication/electron microscopy)

SUSAN BRATOSIN*, ORGAD LAUB*, JACOV TALO, AND YOSEF ALONI**Department of Genetics, The Weizmann Institute of Science, Rehovot, Israel; and tDepartment of Biology, Ben Gurion University of the Negev,Beer Sheva, Israel

Communicated by Michael Sela, June 1, 1979

ABSTRACT During an electron-microscopic survey withthe aim of identifying the parvovirus MVM transcription tem-plate, we observed previously unidentified structures ofMVMDNA in lysates of virus-infected cells. These included double-stranded "lasso"-like structures and relaxed circles. Bothstructures were of unit length MVM DNA, indicating that theywere not intermediates formed during replication; they eachrepresented about 5% of the total nuclear MVM DNA. Theproportion of these structures was unchanged after digestionwith sodium dodecyl sulfate/Pronase and RNase and after milddenaturation treatment. Cleavage of the "lasso" structures withEcoRI restriction endonuclease indicated that the "noose" partof the "lasso" structure is located on the 5' side of the genomicsingle-stranded MVM DNA. A model is presented for the mo-lecular nature of the circularization process of MVM DNA inwhich the "lasso" structures are identified as intermediatesduring circle formation. This model proposes a mechanism forcircularization of linear DNAs.

Animal viruses having linear DNAs provide useful model sys-tems for the study of eukaryotic DNA replication and somaticrearrangement in animal cells. For example, the parvoviruseshave linear single-stranded DNA genomes with an Mr of about1.5 X 106. The genomic DNA contains a stable hairpin duplexstructure at its 5' terminus and a 3'-terminal hairpin structuresuitable for priming complementary strand synthesis in vitro.Both hairpins are believed to be involved in the initiation ofsubsequent cycles of replication, generating complex double-stranded intermediates with a y-shaped appearance, as seen inthe electron microscope (1).

Based on the Cavalier-Smith model for replication of linearDNA (2), Tattersall and Ward (3) have suggested the "rollinghairpin" model to explain how the structures at the ends of thelinear MVM DNA are replicated. The central process in thismodel is the synthesis and rearrangement of the palindromicterminus to produce a "rabbit-eared" structure. Circularstructures are not considered in this model because no circularforms of MVM DNA have been detected in lysates from in-fected cells.The parvoviruses, because of their relative simplicity, also

provide one of the best model systems to date for studyingtranscription and posttranscriptional modifications (4), spe-cifically, the mechanism of splicing (5-7) and the physiologicalsignificance of this process in animal cells.

In a study with the aim of determining the MVM templateof transcription, we have lysed MVM-infected cell nuclei withTriton X-100 and Sarkosyl, a procedure that we have found

suitable for the analysis of simian virus 40 transcriptionalcomplexes (8), and analyzed the lysates by electron micros-copy.We report the occurrence in this lysate of "lasso"-like

structures and circles of MVM DNA. A model is presented forthe molecular nature of the circularization process of MVMDNA in which the "lasso" structures are considered as inter-mediates during circle formation. This model incorporates amechanism for circularization of linear DNAs.

MATERIALS AND METHODSPlaque-purified strain T of MVM was a generous gift of P.Tattersall. Virus propagation and purification were carried outas described (9). A-9 cells, a derivative of the mouse L cell line(10), were grown in 15-cm-diameter petri dishes in Dulbecco'smodified Eagle's medium (GIBCO) with 5% fetal calf serum.Cultures were grown at 370C in an atmosphere containing 10%Co2.Ten petri dishes each containing 6 X 106 A-9 cells were each

infected with 5 ml of a suspension of MVM in phosphate-buf-fered saline (1-3 plaque forming units per cell). After a 30-minperiod at 370C for virus absorption, 20 ml of fresh mediumcontaining 5% fetal calf serum was added to each culture. Ad-ditionally, 0.1 mCi (1 Ci = 3.7 X 1010 becquerels) of [methyl-3H]thymidine (38.4 Ci/mmol, Nuclear Research Center-Negev,Beer-Sheva, Israel) was added to two of the cultures. At 20 hrpostinfection, cells were washed twice with ice-cold phos-phate-buffered saline and treated with Nonidet P-40 detergent(0.5%). The nuclear fraction was isolated (11), and the nucleiwere lysed with 0.3% Triton X-100 and 0.3% Sarkosyl in 0.2 MNaCl/10 mM Tris-HCl, pH 7.4/1 mM EDTA (8). The cellularchromatin was pelleted by centrifugation at 30,000 X g for 20min at 2°C, and the nucleoplasmic supernatant, before or aftertreatment with 0.5% sodium dodecyl sulfate (NaDodSO4)/Pronase at 20 ,tg/ml at 26'C for 30 min, was sedimentedthrough a 5-20% neutral sucrose gradient in 0.2 M NaCl/10mM Tris-HCl, pH 7.4/1 mM EDTA. Pooled fractions from thesucrose gradient were treated with 10 ,ug of pancreatic RNaseper ml at 26°C for 30 min. For control experiments, nucleo-plasmic supernatant was prepared simultaneously from unin-fected cells.

Samples (15 ,ul) from various fractions of the sucrose gradientwere spread for electron microscopy by using either theKleinschmidt aqueous procedure (12) as described by Davis etal. (13) or the urea/formamide technique (14). Grids were ro-tary shadowed with platinum/palladium and examined byusing a Philips 300 electron microscope. Molecules of interest

Abbreviation: NaDodSO4, sodium dodecyl sulfate.4289

The publication costs of this article were defrayed in part by pagecharge payment. This article must therefore be hereby marked "ad-vertisement" in accordance with 18 U. S. C. §1734 solely to indicatethis fact.

4290 Biochemistry: Bratosin et il.

0

E

5 10 I5 20 25Fraction number

FIG. 1. Sedimentation analysis of [3H]DNA extracted fromMVM-infected and mock-infected cells by the Triton/Sarkosyl pro-

cedure. Triton/Sarkosyl lysates were centrifuged through a 5-20%sucrose gradient containing 0.2 M NaCl, 0.01 M Tris.HCl (pH 7.9),and 1 mM EDTA in an SW 65 rotor at 58,000 rpm for 2 hr at 41C.Aliquots (10 Al) of each fraction were assayed for radioactivity inTriton-based scintillation fluid. *, Lysates of infected cells; 0, lysatesof mock-infected cells; A, profile obtained when the Triton/Sarkosyllysate of infected cells was treated with 0.5% NaDodSO4 and 20,4gof Pronase per ml at 260C for 30 min before centrifugation as

above.

were photographed at X40,000 and projected at a final mag-nification of X200,000, and contour lengths were traced andmeasured with a map measurer.

RESULTSElectron Microscopic Analysis of MVM DNA in In-

fected-Cell Lysates. A-9 cells were infected with MVM or

mock-infected and immediately labeled with [3H]thymidine.At 20-hr postinfection, nuclei were isolated (8); labeled DNAwas extracted with Triton X-100 and Sarkosyl and sedimentedthrough sucrose gradients. The radioactive material extractedfrom infected cells was 7-fold that extracted from controlmock-infected cells. Sedimentation profiles are shown in Fig.1. The labeled DNA of infected cells showed a broad band withan average sedimentation coefficient of about 17 S, whereas thelabeled DNA of mock-infected cells showed a smaller band atabout 24 S.

Electron microscopic examination of various fractions of thegradient of labeled DNA from infected cells showed the exis-tence of extended double-stranded DNA and collapsed RNAmolecules. About 90% of the DNA molecules were of unitlength (1.4 jim) MVM DNA (9). The rest were larger molecules,

W'*i.'circles spread by the aqueous technique. Pooled fractions 10-12 (see

legend to Fig. 1) were spread. (A) "Lasso" structures. (B) Relaxed

circle. Bar = 0.5 pm.

FIG. 3. Electron micrographs of "lasso" structures and relaxedcircles spread by the formamide technique. Pooled fractions 10-12(see legend to Fig. 1) were spread. (A) "Lasso" structures. (B) Relaxedcircle. Bar = 0.5,pm.

which could be either dimers and larger forms of MVM DNA(9) or cellular DNA. Surprisingly, about 10% of the unit lengthMVM DNA molecules appeared as "lasso"-like structures andrelaxed circles, forms not previously identified. Although ob-served in the parvovirus H-1 DNA prepared from semiper-missive cells, their significance was not clearly recognized (15).Figs. 2 and 3 show representative DNA molecules spread forelectron microscopy by the aqueous and formamide techiques,respectively. The structures visible in both figures lead us to theconclusion that the molecules are double-stranded over theirentire length. Moreover, the stability of the structures, spreadunder mild denaturing conditions, indicates that there are no

short base-paired regions. The "noose" of the "lasso" structuresvaries in size from very small up to almost full circle. Thecomplete circles were all relaxed and no superhelical circles(form I) (16) were observed. Larger circles having a circum-

ference 3-fold that of the small circles were also seen; becausethey appear also in the lysates of mock-infected cells, they are

probably of cellular origin. Fractions 10-12 of the gradient (Fig.1) were enriched in "lasso"-like structures and relaxed circles,whereas fractions 5-7 were enriched in the large circle struc-tures.

Are Proteins or RNA Involved in Formation of "Lasso"Structures and Relaxed Circles? The Triton/Sarkosyl extractwas treated with NaDodSO4/Pronase or with RNase. It can beseen in Fig. 1 that the NaDodSO4/Pronase treatment has re-

duced the sedimentation rate of the labeled DNA, indicatingthat some proteins were still attached to the DNA extracted bythe Triton/Sarkoskyl method (17). We have scored in theelectronmicrographs linear double-stranded DNA, "lasso"structures, and relaxed circles occurring before and after theNaDodSO4/Pronase treatment. The results, presented in Table1, show that the "lasso" structures and relaxed circles eachrepresent about 5% of the total MVM DNA and that no signif-icant differences can be detected between the treated anduntreated samples. Similar results were obtained for RNase-

Table 1. Occurrence of various forms of unit length MVM DNATotal Molecules scored as:

Extraction molecules Relaxedprocedure scored Linear "Lasso" circles

Triton/Sarkosyl 863 774 48 41(89.6%) (5.6%) (4.8%)

Pronase/NaDodSO4 1268 1129 67 72treatment of (89.0%) (5.3%) (5.7%)Triton/Sarkosyl extract

Fractions 10-12 from the Triton/Sarkosyl extraction procedure andfractions 16-18 from the NaDodSO4/Pronase-treated samples, asshown in Fig. 1, were spread for electron microscopy by the aqueoustechnique. The occurrence of linear unit length MVM DNA, "lasso"structures, and relaxed circles in each preparation was determined.

Proc. Natl. Acad. Sci. USA 76 (1979)

Proc. Natl. Acad. Sci. USA 76 (1979) 4291

treated samples. Because the circular DNA-protein complexof adenovirus is linearized under similar NaDodSO4/Pronasetreatment (18, 19), it appears that, if a protein links the ends ofthe MVM DNA molecule to form the relaxed circles or the"noose" in the "lasso" structures, it should be different fromthat in adenovirus.Are "Lasso" Structures Intermediates during Replication?

The lengths of the "lasso" structures and the circumferencesof the relaxed circles were compared to those of linear dou-ble-stranded MVM DNA and genomic DNA. If the "lasso"structures are replicating intermediates, their lengths shouldincrease with an increase in the proportion of the "noose" lengthin the "lasso" compared to its linear part, reaching a circlehaving a circumference twice as long as that of the originallinear DNA (i.e., 3 gim). In the histograms presented in Fig. 4,the "lasso" structures and relaxed circles show sharp peaks at1.4 ± (SD) 0.1 jgm. The large circles, which do not seem to beMVM-specific (see above), have circumferences of 4.6 + 0.2jim in length. Double-stranded MVM DNA and genomic MVMDNA were found to have the same average lengths (results notshown). Comparison with the length of simian virus 40 DNA(5224-5226 base pairs) (20, 21) leads to a value of 4400 ± 314for the number of base pairs accommodated on all forms- ofdouble-stranded MVM. Similar results for the length of MVMDNA were found by Bourguignon et al. (9). The results pre-sented here, therefore, indicate that the "lasso" structures arenot intermediates during replication.

Is There a Preferred Side on Viral DNA for the "Noose"of the "Lasso"? Electron microscopic analysis of "lasso"structures cleaved with EcoRI restriction endonuclease showedthat cleavage of double-stranded DNA occurred at two sites.These sites were mapped at 20.5 and 31.5% from the corre-sponding 3' and 5' ends of the 'genomic DNA molecule, re-spectively (1). If there is no preferred side for the "noose," thena mixed population of cleaved "lasso" structures, 20.5 and 31.5%of the unit length MVM DNA, would be expected. Fig. 5 showssome typical cleaved "lasso" structures. Lengths of the cleavedstructures are shown in Fig. 6. A histogram with a single peak,corresponding to 33.3 + 1.5% of the whole length of the viralDNA, was obtained, indicating that the "noose" of the "lasso"

Length (Mm)FIG. 4. Histograms showing lengths of "lasso" structures and

relaxed circles. (A) "Lasso" structures as shown in Fig. 3 were selectedat random. Note no accumulation of any specific size of the "noose"part in the "lasso" structures. (B) All the circles measured (Fig. 3B)were relaxed.

FIG. 5. Electron micrographs of "lasso" structures cleaved withEcoRI restriction endonuclease. DNA from pooled fractions 10-12(see legend to Fig. 1) was cleaved with 10 units of EcoRI restrictionendonuclease at 371C for 20 min and spread for electron microscopicanalysis by the aqueous technique.

structure is located mainly at the side corresponding to the 5'end of the genomic DNA.

Molecular Nature of the Circularization Process. The datapresented lead us to suggest that the "lasso" structures representsteps in the circularization process of the linear double-strandedMVM DNA. Fig. 7 presents a model of this process.The first steps (I-IV) proposed for the circularization process

are the same as those in the "rolling hairpin" model (3). Namely,the gap-fill synthesis of the complement d of sequence D (stepII) is followed by displacement synthesis and gap-fill synthesisso that the 5'-terminal hairpin sequence aBA (step III) is copied.The palindrome is now rearranged to a "rabbit-eared" structure(step IV).We suggest that the next step is ligation of the 5' A sequence

and 3' a sequence of the "rabbit-eared" structure (step V). Eachof the "rabbit ears" can then rotate as shown in either step VIor step VI'. Both configurations are presumably not stable,because sequences B and b are not base-paired. A more stableconfiguration, in which sequences B and b are base-paired,would obtain if, instead of base pairing within the palindrome,base pairing occurred between the complementary sequencesof the two palindromes. Steps VII, VII' and VIII, VIII' representintermediates formed during this process. This step can yieldtwo types of "lasso" structures: the ligated Aa sequences caneither be interwrapped between the D and d sequences (stepIX) or they can overwrap sequence Dd (step IX'). Once the"noose" is formed, it can slide towards the opposite end of themolecule (step X and X'). It is noteworthy that an interwrappmginitiation mechanism has been suggested for formation of therecombination intermediates that require no DNA synthesis(22). It is possible, although unlikely, that NaDodSO4/Pro-nase-resistant protein or a nuclear membrane component ispresent at the junction of the "noose" and stabilizes it.

If the interwrapping structure is the correct one, then a

'i 20_IV0

Eo 10_.0

Ez

Ql Q2 0.3 04 Of 06 Q7 0.8 CBMVM fractional length

1.0

FIG. 6. Histogram showing lengths of "lasso" structures aftercleavage with EcoRI restriction endonuclease. Molecules as shownin Fig. 5 were measured. The arrows indicate the two cleavages sitesof EcoRI restriction endonuclease on the corresponding genomicMVM DNA. The left and right ends represent the 5' and 3' ends ofthe genomic DNA, respectively.

] rerte--Of osetrz v] veered row |Il

Biochemistry: Bratosin et al.

tEm OF'rvwm MM

4292 Biochemistry: Bratosin et al.

AI SC

a_

mID EX

aD= Fa D E

a bA d p~~~~~~~~~~~~I__.Mmp

M

A - d ~~e .A d e

D E

-Wff D EBA

b

xI 0~~~~~~F

B A

I F

FIG. 7. Model for circularization of linear DNA. Steps I-IV are

the same as in Tattersall and Ward (3). Steps are as follows: I,Structure of MVM genome. The upper and lower cases denote com-

plementary sequences. The arrow denotes the 3'-OH end of a strand.II, Gap-fill synthesis. III, Gap-fill synthesis to complete genomichairpin. IV, Enzymatic rearrangement of terminal palindrome to give"rabbit-eared" structure. V, Ligation of the Aa sequences of the twopalindromes. VI and VI', Rotation of the two palindromes in the sameor opposite directions, respectively, generating the interwrapping or

overwrapping structures. VII and VII', Intermediates formed during"breathing" of the base pairs in the palindromes. VIII and VIII',Movement of the palindromes. IX and IX', Rearrangement of thecomplementary sequences of the two palindromes to generate the"lasso." X and X', The sliding process. XI, Interlocked relaxedcircle.

protein like the unwinding protein is likely to be involved inthe advancing of the "noose" to yield the interlocked relaxedcircle (see step XI). A further nicking and ligation process couldconvert the interlocked circle to a'continuous double-strandedunit length circle. In the alternative possible mechanism, as

shown in steps IX' and X', a circle is not formed, unless nickingand ligation takes place when sequences Aa are in the proximityof sequences EFe. Because "lasso" structures and circles are

highly enriched only when extraction is carried out in thepresence of Triton, we suggest that if the mechanism of circleformation proceeds as shown in steps IX' and X' then sequences

EFe are attached to a nuclear membrane, which could alsoharbor the nicking and ligating enzymes.

DISCUSSIONIn the present communication, we have reported the occurrence

in MVM-infected cell lysates of different DNA structures ofparvovirus MVM. These include "lasso" structures and relaxedcircles. The MVM origin of the "lasso" structures and circles

is suggested by three observations: (i) they were found in in-fected and not in uninfected cells, (ii) the lengths of the "lasso"structures and circles were found to be the same as that ofgenomic MVM DNA, and (iii) EcoRI restriction enzymecleaved the "lasso" structures at sites typical for MVM DNA.The ability to withstand mild denaturing conditions indicates

that the "noose" must be connected at the fork region in a waythat maintains the continuity of the four polynucleotide chains.The fact that these structures were not destroyed by Na-DodSO4/Pronase and RNase treatments indicates that noprotein nor RNA links the "noose" to the linear part of the"lasso." The growth of the "noose" during sliding, as shown inFig. 7, may involve an unwinding protein (23) acting cooper-atively in melting the double-stranded DNA and orienting thesliding of the single strand. The same unwinding protein canalso reverse the effect of facilitating base pair alignment be-tween the single strands (23). There may be contributions bystill unknown proteins that, together with the unwinding pro-tein, participate in the sliding process. The location of the eu-karyotic unwinding protein in the nuclear membrane, associ-ated with lipids (24), is in agreement with the suggestion thatthe "lasso" structures and relaxed circles are attached to nuclearmembrane (see below). The unwinding protein could still beattached to the "lasso" structures and relaxed circles extractedby the Triton/Sarkosyl method, because some proteins weresubsequently removed by NaDodSO4/Pronase treatment (seeFig. 1).The question of why "lasso" structures and relaxed circles

of MVM DNA have not been observed previously may beconnected with the procedure for preparing infected cell lys-ates. In the present study, the lysates were prepared by usinga combination of the nonionic detergent Triton X-100 and theionic detergent Sarkosyl, instead of the more commonly usedprocedure of Hirt (25) in which the nonionic detergent is notincluded. Indeed, when infected-cell lysates were prepared bythe Hirt method, we were unable to detect any relaxed circlesand found only a very limited number of "lasso" structures. Itis, therefore, suggested that the "lasso" structures and relaxedcircles of MVM DNA may be attached to a nuclear membranethat is lysed by Triton X-100. It is also possible that "lasso"structures and relaxed circles were lost in the Hirt pellet."Lasso"-like structures were observed in infected-cell lysatesof another parvovirus (H-1) (15), prepared by the Hirt proce-dure. However, because these structures appeared to only a verylimited extent, their significance was considered open toquestion (15).

It is noteworthy that no superhelical forms of MVM DNAwere observed in the present study, suggesting that the inter-locking of the double-stranded DNA prevents superhelicity ofthe circle. The failure to observe supercoiled circles may reflectonly their rarity in infected-cell lysates. It should be pointed out,however, that we have not ruled out the unlikely possibility thatthe described structures were produced during the isolation ofthe DNA.

According to the model presented in Fig. 7, linear DNA withpalindromic structures and ligated ends can circularize aftersliding. Because of these limited demands for circularization,it seems quite simple for any linear DNA, chromosomal or viral,like that of the RNA tumor viruses, for example, to circularize.Indeed, eukaryotic cells often contain populations of circularDNAs that derive at least in part from chromosomal DNA (26).Moreover, a precursor-product relationship has been demon-strated between cytoplasmic linear and covalently closed nu-clear forms of avian sarcoma virus DNAs (27). In both cases, themolecular nature of the circularization process has not beencharacterized and no biological function has been assigned to

Proc. Natl. Acad. Sci. USA 76 (1979)

zME

A So' D Eb

d aly

D E

Proc. Natl. Acad. Sci. USA 76 (1979) 4293

any of these circular DNAs. It has been speculated, however,that the eukaryotic circular DNA might be the product of generearrangement, whereas the viral circular DNA may serve asthe immediate precursor for integration.The reason for circularization of the unit length double-

stranded MVM DNA is not known at the present time. Circu-larization could be a necessary step for integration, for repli-cation, or for providing a suitable template for transcription.

This research was supported by part by U.S. Public Health ServiceResearch Grant CA 14995.

1. Ward, D. C. & Dadachanji, K. (1978) in Replication of Mam-malian Parvoviruses, eds. Ward, D. C. & Tattersall, P. (ColdSpring Harbor Laboratory, Cold Spring Harbor, NY), pp.297-313.

2. Cavalier-Smith, T. (1974) Nature (London) 250, 467-470.3. Tattersall, P. & Ward, D. C. (1976) Nature (London) 263,

106-109.4. Tal, J., Ron, D., Tattersall, P., Bratosin, S. & Aloni, Y. (1979)

Nature (London) 279, 649-651.5. Berget, S. M., Moore, C. & Sharp, P. A. (1977) Proc. Natl. Acad.

Sci. USA 74,3171-3175.6. Aloni, Y., Dhar, R., Laub, O., Horowitz, M. & Khoury, G. (1977)

Proc. Natl. Acad. Sci. USA 74,3686-3690.7. Chow, L. T., Gelinas, R. E., Broker, T. R. & Roberts, R. J. (1977)

Cell 12, 1-8.8. Laub, O., Bratosin, S., Horowitz, M. & Aloni, Y. (1979) Virology

92,310-323.9. Bourguignon, G. J., Tattersall, R. J. & Ward, D. C. (1976) J. Virol.

20,290-306.

10. Littlefield, J. W. (1964) Nature (London) 203; 1142-1144.11. Penman, S. (1966) J. Mol. Biol. 17, 117-130.12. Kleinschmidt, A. K. & Zahn, R. K. (1959) Z. Naturforsch. Teil

B 14, 770-779.13. Davis, R. W., Simon, M. N. & Davidson, N. (1971) Methods

Enzymol. 21, 413-428.14. Bratosin, S., Horowitz, M., Laub, 0. & Aloni, Y. (1978) Cell 13,

783-790.15. Rhodes, S. L., III (1978) in Replication of Mammalian Parvo-

viruses, eds. Ward, D. C. & Tattersall, P. (Cold Spring HarborLaboratory, Cold Spring Harbor, NY), pp. 279-295.

16. Radloff, R., Bauer, W. & Vinograd, J. (1967) Proc. Natl. Acad.Sci. USA 57, 1514-1521.

17. Laub, 0. & Aloni, Y. (1976) Virology 75,346-354.18. Robinson, A. J., Younghusband, H. B. & Bellet, A. J. D. (1973)

Virology 56,54-69.19. Rekosh, D. M. K., Russell, W. C., Bellet, A. J. D. & Robinson, A.

J. (1977) Cell 11, 283-295.20. Reddy, V. B., Thimmapaya, B., Dhar, R., Subramanian, K. N.,

Zain, B. S., Pan, J., Ghosh, P. K., Celma, M. L. & Weissman, S.M. (1978) Science 200, 494-502.

21. Fiers, W., Contreras, R., Haegeman, G., Rogers, R., Van deVoorde, A., Van Heuverswyn, H., Van Herreweghe, J., Volekaert,G. & Ysebaert, M. (1978) Nature (London) 273, 113-120.

22. Potter, H.'& Dressier, D. (1978) Proc. Natl. Acad. Sci. USA 75,3698-3702.

23. Sigal, N., Delius, H., Kornberg, T., Gefter, M. L. & Alberts, B.(1972) Proc. Natl. Acad. Sci. USA 69, 3537-3541.

24. Hotta, Y. & Stern, H. (1971) Dev. Biol. 26, 87-99.25. Hirt, B. (1967) J. Mol. Biol. 26,365-369.26. Smith, C. A. & Vinograd, J. (1972) J. Mol. Biol. 69, 163-178.27. Shank, P. R. & Varmus, H. E. (1978) J. Virol. 25, 104-114.

Biochemistry: Bratosin et al.