`Vol. 74, N0. 10, pp. 4168-4172, October 1977
`Biochemistry
`
`On the mechanism of genetic recombination:
`The maturation of recombination intermediates
`
`(plasmid DNA/chi form/Holliday recombination intermediate/DNA multimers/recA gene/intramolecular recombination)
`
`HUNTINGTON POTTER AND DAVID DRESSLER
`
`Department of Biochemistry and Molecular Biology, Harvard University, Cambridge, Massachusetts 02138
`
`Communicated by J. D. Watson, June 23, 1977
`
`,
`
`DNA molecules of the plasmid ColEl are
`ABSTRACT
`normall
`recovered from wild-type cells as a set of monomer-
`and mu timer-size rings. The data of this paper show that the
`multimer-size species are a product of enetic recombination.
`Multimer rings do not arise after trans ection of purified mo-
`nomers into bacterial host cells lacking a functional recA re-
`combination system. Analogously, purified dimers, trimers, and
`tetramers, transfected into recA ‘ cells, can replicate, but are
`constrained to remain in those conformations. Only upon
`transfection into rec+ cells can they regenerate the full spectrum
`of monomer- and multimer-size species.
`In this paper we trace the flow of genetic information from
`the monomer to the multimer state and back again under the
`guidance of the teat recombination system.
`The formation of multimer-size DNA rings is discussed as a
`natural consequence of the maturation of a Holliday recombi-
`nation intermediate formed between two monomer plasmid
`genomes.
`'
`
`Controlled genetic matings in eukaryotes have led to a wealth
`of information about the process of genetic recombination.
`From this information Holliday in 1964 suggested a molecular
`mechanism for recombination (1-3). The cornerstone of this
`mechanism is a central recombination intermediate that can
`be formed and matured to account for four genetic findings
`about recombinant eukaryotic chromosomes. These are: (i) All
`of the chromosomes that enter a eukaryotic meiosis and par-
`ticipate in recombination are recovered. That is, the recombi-
`nation process is carried out with a net conservation of genetic
`information (4). (ii) The exchange of genetic information be-
`tween homologous chromosomes (as observed over distances
`longer than a few thousand nucleotides) is almost always pre-
`cisely reciprocal (4). (iii) An area of heterozygous DNA is fre-
`quently fonned in the immediate region of the recombination
`event (5-10). (iv) The recombining chromosomes are appar-
`ently broken apart so that there is an equal chance that the genes
`on either side of the crossover position (the region marked by
`potential heterozygosity) will either be left in their parental
`linkage or emerge in a recombinant linkage (5, 6).
`The Holliday recombination intermediate, which can ac-
`count for these findings, is shown in Fig. 1. The intermediate
`is most easily thought of as arising from the pairing of two ho-
`mologous DNA molecules that become nicked at roughly
`equivalent places and undergo a reciprocal strand invasion (Fig.
`1 A—C). This sequence of events, or a related sequence (11), is
`expected to result in the formation of a stable recombination
`intermediate that, for convenience, can be represented in either
`of two planar configurations (Fig. 1 F and G). The subsequent
`maturation of the Holliday intermediate to yield recombinant
`chromosomes is shown in Fig. 1 H and I .
`
`The costs of publication of this article were defrayed in part by the
`payment of page charges. This article must therefore be hereby marked
`“advertisement" in accordance with 18 U. S. C. §l734 solely to indicate
`this fact.
`
`While the Holliday model was evolved from a consideration
`of the genetic composition of recombinant chromosomes in
`eukaryotes, more recent evidence indicates that this mechanism
`may also be applicable to recombination in prokaryotes. Im-
`proved procedures for isolating nucleic acids and the intro-
`duction of electron microscopy (which allows the study of in-
`dividual DNA molecules) have provided support for the H01-
`liday intermediate in several prokaryotic systems; basically
`these relate to the recombination system of Escherichia call as
`studied through its viruses S13, ¢X174, and A (12-15), and
`through the DNA of one of its plasmids, OolE1 (16, 17). In some
`of these instances it has been possible to directly visualize the
`recombination intermediate proposed by Holliday.
`In our work we use plasmid DNA as an experimental system
`for studying recombination intermediates-—both because the
`plasmid DNA molecules are small and can easily be isolated
`from the cell without breakage, and because the number of
`plasmid molecules can be selectively amplified to about 1000
`_per cell, potentially enhancing the opportunity for recombi-
`nation by increasing the number of homologous DNA se-
`quences. As discussed in a previous paper, we have been able
`to recover dimeric plasmid DNA molecules shaped like figure
`8s, consistent with the interpretation of two circular genomes
`interacting at a point of DNA homology (see Fig. 2A, and also
`refs. 16 and 17). We confirmed this interpretation by linearizing
`the genomes of the plasmid with the restriction enzyme EcoRI:
`the figure-8 structures were converted to molecules with bi-
`lateral symmetry, shaped like the Greek letter chi (x) (an ex-
`ample is shown in Fig. 2B). Over 1000 such “x forms” were
`photographed and analyzed. Each molecule contained two pairs
`of equal-length arms, indicating that two plasmid genomes
`were held together at a point of DNA homology. Moreover, in
`about 100 instances it proved possible to observe the polynu-
`cleotide strands in the region of the crossover. These strands
`could be seen connecting the two genomes, crossing over from
`one to the other (as in Fig. 2C).
`The involvement of the x forms in recombination was in-
`ferred from their absence in recA ’ cells.
`We have interpreted our results as offering physical evidence
`in support of the recombination intermediate postulated by
`Holliday on genetic grounds: the experimentally observed x
`forms correspond exactly to the two planar representations of
`the Holliday intermediate (compare Fig. 2 B and C with Fig.
`1 F and C).
`The physical evidence in support of the Holliday recombi-
`nation pattern, such as shown in Fig. 2, deals specifically with
`the structure of the recombination intermediate. Two aspects
`of the recombination process that are less well understood
`concern the initiation events leading to the formation of the
`intermediate and the maturation steps by which the interme-
`diate yields recombinant chromosomes This paper is concerned
`with the maturation stage of the recombination process.
`
`4168
`
`Mylan v. Genentech
`Mylan V. Genentech
`IPR2016-00710
`Genentech Exhibit 2066
`
`IPR2016-00710
`
`Genentech Exhibit 2066
`
`
`
`Biochemistry: Potter and Dressler
`
`Proc. Natl. Acad. Sci. USA 74 (1977)
`
`4169
`
`A
`
`Z
`
`A
`
`a
`
`—;:—~—-
`z
`
`Z’
`_
`+
`
`B
`
`
`
`
`
`
`
`
`
`A
`
`ZA
`
`2
`
`a
`
`za
`
`Z
`
`FIG. 1. Linear DNA molecules undergoing genetic recombination via a Holliday intermediate. The model is explained in the text.
`
`The maturation problem
`
`One of the most interesting features of the Holliday interme-
`diate arises because of its inherent symmetry. This symmetry
`leads to the expectation that the intennediate can be processed
`in either of two related ways to yield two different pairs of re-
`combinant chromosomes. This dual maturation potential is most
`easily visualized if one considers the recombination interme-
`diate in the planar reprmentation shown in Fig. 1G. One of the
`maturation pathways cuts diagonally across the intermediate
`
`FIG. 2. Electron micrographs of plasmid DNA molecules in the
`process of recombination. (A) A figure-8 structure containing two
`monomer-size plasmid genomes. (B) A chi form: the plasmid DNA
`has been linearized with the restriction enzyme EcoRI (as in ref. 17);
`the two plasmid genomes are seen to be interacting at a region of DNA
`homology (compare with Fig. 1F). (C) An “open” x form: during
`spreading for the electron microscope, the double helix has been
`disrupted in the region of the crossover, allowing the strand sub-
`structure to be seen (compare with Fig. 1G). Following purification
`from recA+ cells, the plasmid DNA was examined in the electron
`microscope by a modification (18) of the Kleinschmidt and Zahn
`protein monolayer technique (19) as previously described (20).
`
`along a north—south axis and leads to the formation of a pair of
`reciprocally recombinant chromosomes in which the genes on
`either side of the crossover emerge in a_ newlinkage. This is a
`traditional single recombination event. But due to the symmetry
`of
`the recombination intermediate, a second maturation
`pathway is equally likely. In this case, diagonal nicking along
`an east—west axis leads to the separation of the two recombining
`chromosomes—but this time the genes on either side of the
`crossover are left in their original linkage. From a genetic point
`of view the recombination event would be silent, were it not for
`the possibility of forming heterozygous DNA in the immediate
`region of the crossover (as in Fig. 1 D -> E).
`The maturation process, as discussed so far, relates to linear
`DNA molecules such as those found in eukaryotic chromosomes.
`The chromosomes of prokaryotes, viruses, and cellular organ-
`elles are, on the other hand, generally circular. It has long been
`recognized that this circularity poses a problem for the re-
`combination process: a single recombination event between
`circular chromosomes is expected to produce a composite
`multimer structure. This sequence of events can be readily seen
`if one applies the processing steps we have considered to the
`maturation of the Holliday intermediate formed between two
`circles. As shown in Fig. 3, one of the two possible maturation
`pathways cuts the figure-8 intermediate apart into two inde-
`pendent monomer rings. But in these monomers, the outside
`markers are always left in their original linkage, and the only
`genetic recombination that arises does so because of the possible
`formation of heterozygous DNA in the immediate region of the
`crossover. On the other hand, if the alternative maturation nicks
`are introduced, the recombination intermediate is automatically
`converted into a dimer—size circle (Fig. 3C).
`Thus monomer rings with a recombinant arrangement of the
`genes on either side of the crossover point are_ not directly
`produced by maturation of the recombination intermediate
`formed between two circles. It is the purpose of this paper to
`present evidence bearing on this problem.
`Multimers are formed in recA+ cells
`
`As discussed in the previous section, it is expected that circular
`dimers are a product of the maturation of two monomer circles
`engaged in the crossover stage of genetic recombination (see
`Fig. 3).
`We have explicitly tested this prediction by purifying mo-
`nomer-size plasmid rings and introducing them into rec+ and
`recA' cells. The transplantation of the DNA rings into the
`
`
`
`4170
`
`Biochemistry: Potter and Dressler
`
`Proc. Natl. Acad. Sci. (lSA 74 (1977)
`
`
`
`FIG. 3. Circular DNA molecules undergoing recombination via a Holliday intermediate. Because of the symmetry inherent in the recombi-
`‘ nation intermediate (E) there are two alternative maturation pathways. One leads to the -formation of two monomer rings with the genes on
`either side of the crossover position retaining their original linkage. The other maturation pathway leads to the formation of a circular dimer.
`
`desired experimental backgroundis readily achieved by using
`recipient cells that have been made temporarily permeable to
`exogenous DNA by calcium treatment and a heat shock. Cells‘
`that receive a plasmid can be selected because they simulta-
`neously gain a tetracycline resistance gene carried on the
`plasmid, and thus acquire the ability to form a colony on solid
`medium supplemented with tetracycline. Each colony‘ that
`appears contains the entire lineage of plasmids resulting from
`the entrance of a single DNA molecule into a single recipient
`cell.
`Several tetracycline-resistant colonies were picked from
`among the 1ec+ and recA ' recipients of purified plasmid DNA
`monomers. Each colony was further grown in the presence of
`tetracycline in logarithmic-phase culture for 70 generations.
`This was done to allow time for the cellular plasmid population
`to approach equilibrium. Plasmid DNA was then extracted
`from the rec+ and recA" cells and analyzed for size hetero-
`geneity using agarose gel electrophoresis.
`As shown in Fig. 4, the monomers transfected into recA ' cells
`were only able to replicate to form more monomers. In contrast,
`the transfection of monomers into rec + cells led to the devel-
`opment of a population containing both monomer and mul-
`timer plasmid genomes. This finding was further solidified by
`analysis using the electron microscope, which is of higher res-
`olution. Table 1 shows the percentages of monomers, dimers,
`trimers, and tetramers found in rec+ and recA" cells after
`transfection with purified monomers. In recA' cells 99% of the
`plasmids were monomers; only 1% were dimers. In rec+ cells,
`after 70 generations, there were 60% monomers, 37% dimers,
`1% trimers, and 2% tetramers.
`The general conclusion is that in the absence of a functional
`recA system, multimeric plasmid DNA forms do not arise. Or,
`stated conversely, multimeric plasmid genomes appear to be
`the result of recombination.
`'
`
`The action of the recombination system on multimers
`
`As considered so far, the recombination system has produced
`a multimer-size plasmid instead of a pair of monomers that are
`recombinant for the genes on either side of the crossover. For
`the reciprocal recombination process to be completed, there
`’ must be a mechanism for the conversion of the larger circles
`back to the smaller ones. If not to achieve reciprocal recombi-
`nation, such a mechanism would still be necessary to prevent
`multimer-size genomes from progressively accumulating and,
`as the result of recombination, dominating the cell population.
`
`The mechanism for regenerating monomers is not likely to be
`DNA replication because, as shown in Table 1, dimers, trimers,
`and tetramers transfected into recA' cells are locked into those
`configurations and can only replicate to form more dimers,
`trimers, and tetramers, respectively.
`The experiments we will now discuss show that monomer
`genomes can in fact be obtained from a multimer-size plasmid.
`The key is to realize that the multimer circles that are produced
`as a result of the maturation of the first crossover need not be
`end products: they may themselves be substrates for the re-
`combination system.
`If two homologous areas within a dimer DNA ring can ini-
`tiate an intramolecular recombination event, a new figure-8
`intermediate will becreated, atla new crossover position (see Fig.
`5). This intermediate will be identical in strand substructure
`to the figure 8 that is formed by reciprocal strand invasion be-
`tween two monomer rings (compare Fig. 5 A—D with Fig. 3
`A—D). The maturation of this figure 8 should, 50% of the time,
`yield two monomer rings with recombinant arrangements of
`genetic markers (Fig. 5H).
`Thus, we sought to demonstrate that, when dimeric circles
`are placed in rec + cells, the circles are able to undergo an in-
`tramolecular recombination event leading back to the forma-
`tion of monomers. To this end we transfected purified dimers
`
`
`
`(A) Both monomer- and multimer-size plasmid DNA
`FIG. 4.
`rings are recovered after transfection of monomers into rec*' cells. (B)
`Monomers transfected into real‘ cells are able to replicate, but
`cannot recombine to form multimeric structures. To obtain this result,
`plasmid DNA was recovered from individual rec+ and recA‘ clones
`and analyzed for size heterogeneity by agarose gel electrophoresis.
`The gel contained 0.75% agarose in 40 mM Tris.-HCI/20 mM NaOAc/1
`mM EDTA, pH 7.6; after electrophoresis, the DNA was visualized by
`staining with ethidium bromide (0.5 pg/ml for 60 min). The major
`bands, from right to left, are supercoiled plasmid monomers, dimers,
`trimers, and tetramers. The faint bands result from a low percentage
`of nonsuperhelical rings.
`
`
`
`Biochemistry: Potter and Dressler
`
`Proc. Natl. Acad. Sci. USA 74 (1977)
`
`4171
`
`Table 1. Fate of purified monomers, dimers, trimers, and tetramers transfected into rec"’ and recA‘ cells
`
`Transfection
`Resulting plasmid population
`
`DNA
`Recipient
`Monomers
`Dimers
`Trimers
`Tetramers
`Higher multimers
`Total
`Monomers
`rec+
`595
`367
`13
`24
`1
`1000
`recA ‘
`994
`6
`0
`0
`0
`1000
`
`Dimers
`
`~ Trimers
`
`Tetramers
`
`rec +
`recA ‘
`
`rec +
`recA ‘
`
`rec*
`recA ‘
`
`43
`1
`
`167
`1
`
`68
`2
`
`693
`989
`
`136
`0
`
`432
`10
`
`8
`0
`
`675
`984
`
`62
`2
`
`241
`9
`
`5
`O
`
`407
`967
`
`15
`1
`
`17
`15
`
`31
`19
`
`1000
`1000
`
`1000
`10()0
`
`1000
`1000
`
`Monomer and multimer plasmid DNA rings were purified by repeated sucrose velocity gradient centrifugation, passaged through recA " cells,
`and then used to transfect, in parallel, rec+ and recA ‘ cells (E. coli strains 294 and 152 from M. Meselson). Individual cells that received a plasmid
`were selected immediately after transfection by virtue of their conversion to a tetracycline-resistant state. From each transfection 2-6 recipient
`cells were recovered and grown into cultures, from which superhelical and nonsuperhelical plasmid DNA was purified (17, 21). The numbers
`of monomer and multimer plasmid species were then determined using the electron microscope. Plasmid sizes were easy to distinguish by in-
`spection; these estimates were confirmed by selective photography and molecule measurement. The experiment has been done four times.
`
`into rec+ cells, selected individual tetracycline-resistant re-
`cipients, grew these cells through 70 generations, and then
`isolated their plasmid DNA.
`As shown in Table 1, we observed that the dimers were
`processed by the recA recombination system to generate both
`larger and smaller plasmid DNA rings. After 70 generations,
`the percentages of monomers, dimers, trimers, and tetramers
`in rec+ cells were 4%, 69%, 1%, and 24%. Similarly, purified
`trimers and tetramers transfected into rec+ cells regenerated
`the full spectrum of monomer and multimer-size rings.
`In sum, these results indicate that recA -mediated intramo-
`
`lecular recombination can occur in DNA circles, allowing the
`production of pairs of reciprocally recombinant genomes.
`_
`Our results may have relevance to a model recently proposed
`by Holloman et al. (22). They have made the interesting pro-
`posal that the initiation of recA-mediated recombination is an
`inherently asymmetric event. Specifically, one recombining
`DNA molecule is supercoiled and the other relaxed, so that the
`former (by giving up its superhelical twists and acquiring a
`region of local denaturation) may serve as the recipient for a
`strand invasion from the latter (22). Under such a model, a
`circular dimer would be expected to be unable to initiate in-
`
`A
`
`FIG. 5. Steps by which a circular multimer can undergo an intramolecular recombination event. The aligning of homologous DNA sequences
`is followed by the nicking and exchange of DNA strands of like polarity, leading to the formation of a figure-8 structure. The strand substructure
`in the crossover region is identical to that of the recombination intermediates shown in Figs. 1 and 3. The two alternatives for maturation of
`this recombination intermediate allow: (1') the re-formation of the multimer (G -r A), or (ii) the creation of a pair of reciprocally recombinant
`DNA rings (G’ —> H).
`
`
`
`
`
`4172
`
`Biochemistry: Potter and Dressler
`
`Proc. Natl. Acad. Sci. USA 74 (1977)
`
`tramolecular recombination because it would be either entirely
`supercoiled or entirely relaxed. Our results are at variance with
`this hypothesis, for we see the production of smaller circles
`when supercoiled multimers are placed in rec+ cells. Perhaps
`a reconciliation is possible if the circular DNA molecules are
`able to possess independently superhelical regions. For instance,
`there could be a cellular system that maintains part of the DNA
`molecule in a superhelical state, while allowing other parts to
`remain relaxed. Independent superhelical regions have been
`observed in the E. colt chromosome (23), but their relevance
`to recombination has not been investigated.
`
`Interpretation
`
`Overall, our evidence supports reciprocal recombination be-
`tween DNA circles as a two-stage process leading first in a bi-
`molecular reaction to a multimer-size structure. Then, a second,
`intramolecular, recombination event breaks the multimer apart
`into two smaller circles in which the original genes have been
`recombined into a new configuration. The data we have pre-
`sented in this paper trace the flow of genetic information from
`the monomer to the multimer state and back again under the
`guidance of the recA recombination system.
`There is a potentially important consequence of the finding
`that multimer-size DNA rings can be recombined into smaller
`circles, for this is essentially a deletion process. Accordingly, it
`would appear that the recA recombination system can work on
`tandem homologies whenever they occur in chromosomal DNA
`to catalyze the formation of a recombination intermediate (as
`in Fig. 5) and the excision of the DNA between the regions of
`tandem homology. How much difficulty this presents for the
`cells is unclear.
`Because intramolecular recombination can occur (Table 1),
`one might be concerned that such control elements as pro-
`moters, which necessarily occur many times on a chromosome,
`would provide a source of repeated homologies inviting cata-
`strophic deletions. What defense is there against such deletions?
`The simplest defense would result if the regions of homology
`were sufficiently small so that recombination events involving
`them would be rare. Also, subtleties in the design of such re-
`peated structures as promoters may be able to prevent recom-
`bination. For instance, initial DNA sequencing studies indicate
`that promoters have been built so as to contain short recognition
`regions interrupted by nonhomologous nucleotide stretches.
`That is, although the RNA polymerase may sit on a 40- to 60-
`base-pair stretch of DNA, it makes contact with only a few key
`base pairs that actually constitute the promoter. The rest of the
`binding area appears to be nonhomologous material, perhaps
`designed to prevent recombination. Thus, the polymerase can
`recognize many promoters as being identical for the purpose
`of binding, but the promoters cannot recognize each other as
`being identical for the purpose of recombination.
`The occurrence of whole genes repeated in tandem, such as
`ribosomal RNA cistrons, presents a more serious problem. Here
`design aspects minimizing the amount of DNA homology are
`not in evidence. Unless a mechanism exists for keeping some
`regions of the chromosome silent for the purpose of recombi-
`nation, we would have to expect that intramolecular recom-
`bination involving these regions would result in a deletion that
`could prove lethal to the occasional cell in which it occurred.
`Two related studies
`
`In a study similar to ours, but based on agarose gel analysis,
`Bedbrook and Ausubel (24) have also determined that mul-
`
`tirneric plasmid genomes are a result of recombination. Where
`they overlap, our two sets of data reinforce each other.
`Another study bearing on the recombinational origin of
`multimer DNA rings involves phage A. A genomes that contain
`mutations in genes gamma and red cannot synthesize conca-
`temeric DNA characteristic of the late life cycle, nor can they
`engage in high efficiency phage-mediated recombination. Such
`phage, in attempting to grow, synthesize predominantly mo-
`nomer-size rings. These rings cannot be packaged in coat pro-
`tein, and the phage form small plaques. Lam et at. (10) have
`studied a class of second-site mutations that restore active phage
`production in this system. They have shown that these muta-
`tions, called x mutations, enhance recombination 20- to 30-fold
`and have hypothesized that the restoration of normal phage
`growth results from an increase in the recombination-mediated
`production of packageable multimer DNA forms.
`We thank Drs. Frank Stahl, Charles Radding, and John Wclfson for
`their careful readings of the manuscript. This study has been funded
`by Research Grants NP-57 from the American Cancer Society, GM-
`17088 from the National Institutes of Health, and the Research Career
`Development Award GM-70440 from the National Institutes of
`Health.
`
`1. Holliday, R. (1964) Genet. Res. 5, 282-304.
`2. Holliday, R. (1968) in Replication and Recombination of Genetic
`Material, eds. Peacock, W. & Brock, R. (Australian Acad. Sci.,
`Canberra), pp. 157-174.
`3. Holliday, R. (1974) Genetics 78, 273-287.
`4. Pascher, A. (1918) Ber. Dtseh. Bot. Ges. 36, 163-168.
`5. Kitani, Y., Olive, L. & El-ani, A. (1962) Am.
`Bot. 49, 697-
`706.
`_
`6. Hurst, D., Fogel, S. & Mortimer, R. (1972) Proe. Natl. Aead. Sci.
`USA 69, 101-105.
`7. Russo, V. (1973) Mol. Gen. Genet. 122,353—366.
`8. White, B. & Fox, M. (1974) Proc. Natl. Acad. Sci. USA 71,
`1544-1548.
`9. Enea, V. & Zinder, N. (1976) Mol. Biol. 101, 25-38.
`10. Lam, S., Stahl, M., McMilin, K. & Stahl, F. (1974) Genetics 77,
`425-433.
`11. Meselson, M. & Radding, C. (1975) Proc. Natl. Acad. Sci. USA
`72, 358-361.
`12. Thompson, B., Escarmis, C., Parker, B., Slater, W., Doniger, ].,
`Tessman, I. & Warner, R. (1975) J. Mol. Biol. 91, 409-419.
`13. Benbow, R., Zuccarelli, A., Shafer, A. & Sinsheimer, R. (1974)
`in Mechanisms in Recombination, ed. Grell, R. (Plenum Press,
`New York), pp. 3-18.
`14. Valenzuela, M. & Inman, R. (1975) Proe. Natl. Acad. Sci. USA
`72, 3024-3028.
`15. Sarthy, V. & Meselson, M. (1976) Proc. Natl. Acad. Sci. USA 73,
`4613-4617.
`16. Potter, H. & Dressler, D. (1976) Proe. Natl. Acad. Sci. USA 73,
`3000-3004.
`17. Potter, H. & Dressler, D. (1977) in DNA Insertion Elements,
`Plasmids, and Eptsomes, eds. Bukhari, A., Adhya, S. & Shapiro,
`I. (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY),
`pp. 409-421.
`18. Davis, R., Simon, M. & Davidson, N. (1971) in Methods in En-
`zymology, eds. Grossman, L. '8} Moldave, K. (Academic Press,
`New York), Vol. 21, pp. 413-428.
`19. Kleinschmidt, A. & Zahn, R. (1959) Z. Naturforsch. Tetl B 14,
`770-781.
`20. Wolfson, ]., Dresler, D. & Magazin, M. (1972) Proc. Natl. Aead.
`Sci. USA 69, 499-504.
`21. Clewell, D. (1972) Bacterial. 110, 667.
`22. Holloman, W., Wiegand, R., Hoessli, C. & Radding, C. (1975)
`Proc. Natl. Acad. Sci. USA 72, 2394-2398.
`23. Worcel, A. & Burgi, E. (1972) Mol. Biol. 71, 127-147.
`24. Bedbrook,
`& Ausubel, F. (1976) Cell 9, 707-716.