- Email: [email protected]
Replication of plasmid R1: Meselson-Stahl density shift experiments revisited
PLASMID
9, 218-221 (1983)
Replication of Plasmid RI : Meselson-Stahl Density Shift Experiments Revisited KURT NORDSTROM’ Department of Molecular Biology, Odense University, Campusvej 55, DK-5230 Odense, Denmark Received April 5, 1982; revised November
16, 1982
Meselson-Stahl density shift experiments have been used extensively to study selection and timing of plasmid replication. Experiments with plasmid RI were previously performed and the conclusion was that this plasmid replicates one copy at a time and that there is an eclipse period after each replication during which no further replications can take place in the cell (Nordstrom et al., Plasmid 1, 187-203 (1978)). However, this interpretation is in conflict with other data, mainly with those obtained in copy number shift experiments (Gustafsson and Nordstrom, J. Bacterial. 141, 106-I 10 (1980)). However, the density shift experiments have now been reinterpreted such that there no longer is any conflict with the copy number shift experiments. There does not seem to be any such eclipse period, but newly replicated plasmid molecules are not available for a second replication for about 20% of a generation time.
Many plasmids are present in just a few (two-five) copies per bacterial cell. Nevertheless, such plasmids are very rarely lost. This demands the existance of replication control mechanisms, a prediction that has proven correct for a large number of plasmids. The plasmids themselves carry genes for control of their own replication (I). One property of a replication control system is that it should be able to correct deviations from the normal copy number. Therefore, a constitutively acting system analogous to a promoter system would not do. Pritchard (2,3) proposed his so-called inhibitor-dilution model, which among other things fulfils this correction requirement. The key element is an inhibitor that is formed in proportion to the concentration of its structural gene, i.e., in proportion to the concentration of the plasmid. Another key assumption of the model is that small fluctuations in inhibitor concentration result in large changes in replication probability. Replication inhibitors have been found for several plasmids. In some cases,the evidence is purely genetic (I); in others, the ’ Present addre~ Department of Microbiology, University of Uppsala, Biomedical Center, P.O. Box 581, S 751 23 Uppsala, Sweden. 0147-619X/83
$3.00
Copyright 0 1983 by Academic Press. Inc. All rights of reproduction in any form resews.
218
inhibitors have been identified as proteins Xdv (4.5) and R 1 (6)) or small RNA molecules (ColEl (7-9) and Rl (10)). Replication of plasmids seemsto be spread over the whole cell cycle (II) (with the possible exception of plasmid F, where the data are conflicting) (I I, 12). Furthermore, plasmid copies are selected randomly for replication; this has been demonstrated for many plasmids by Meselson-Stahl (13) density shift experiments (1 I, 14-16). Such experiments with plasmid Rl have been reported before (II). The following was concluded: (i) plasmid R 1 replicates one copy at a time; (iz) after each replication, there is an eclipse period during which no further replications can occur in the cell; (iii) the newly replicated molecules have the same probability as the other ones in the cell to be selected for the next replication (random selection); and (iv) plasmid replication is spread over the whole cell cycle. However, an increasing body of data obtained with FII plasmids (to which group Rl belongs) is in contradiction to the inhibitor dilution model. Plasmid R 100 (closely related to Rl) replicates with a significant rate even in the presenceof copy mutants of plasmid R 1, which indicates a sloppy responseto high concentrations of inhibitor (I 7). Cloned
SHORT COMMUNICATIONS I
rl
219
inhibitor dilution model (2,3), which predicts a period with total switch-off of plasmid replication in shifts from a high to a lower copy number and a burst of replication in the op posite shift. However, if the data contradict the idea that fluctuations in inhibitor concentration cause drastic changes in replication probability, an eclipse period after each replication ,' I OK&, cannot be explained by fluctuation in inhib1 0.5 0 itor concentration. It may even be possible Generatmns after density shkft that the whole idea of an eclipse period is FIG. 1. Appearance of LLplasmid DNA in an expowrong. Therefore, it may be appropriate to nentially growing population of bacteria during a density shift from heavy (H) to light (L) glycerol-minimal me- have a second look at the interpretation of experiment with respect dium (from Gustafsson et al. (II)). In curve A, the rate the Me&on-Stahl of formation of LLDNA and HLDNA is supposed to to plasmid replication. be=solely a function of the concentrations of LLDNA, In a density shift experiment, plasmid-carHL-DNA, and HH-DNA, which gives a relative amount tying bacteria are grown exponentially in a of HL-DNA = 2(eekL- eezk’)and of LL-DNA = I - 2eSk’ dense medium, e.g., “NH.++, D20, and then + e-*4 where k is the growth constant and t is time. In curve B, in addition to the assumptions made for curve shifted to light medium, e.g., 14NH4+,H20. A, there is a delay time (t,,)of 0.22 generation times, which The bacteria are labeled with [i4C]thymidine makes the kinetics dependent upon the previous history before and with [3H]thymidine after the shift. of the system. The experimentally found values for plasThe plasmid DNA is separated from chromid R I in Escherichia coli K- 12 strain EC 1005have been mosomal DNA and then analyzed for heavy inserted as open circles. (HH)-,* hybrid (HL)-, and light (LL)-DNA by fragments of the inhibitor gene do not com- CsCl density gradient centrif&ation (J I, Z4pletely switch off plasmid R 1 replication (18). 16). The appearance of LGDNA indicates Most important, though, is the result of copy plasmids that have replicated at least twice. For all plasmids studied, the result is banumber shift experiments. The copy number sically the same (plasmid F may be an excepof plasmid RI varies with the growth rate of tion (Rownd, personal communication)). Afthe host bacteria (19). Hence, it is possible to ter one generation, about % of the DNA is perform shift between different copy numbers HH, 3 HL, and ‘/5 LL, indicating random seby shifting the growth conditions (20). The lection for replication. The result of a density result from these experiments was that, irreshift experiment with plasmid RI is shown in spective of the copy number, on the average Fig. 1 (I 1). LGDNA started to appear after n copies are replicated per cell and cell cycle. about 0.3 generations. The data were interHence, in shifts from a high copy number to preted in the following way. There are three a lower one, the replication probability per reactions: cell remains unchanged, whereas the replication probability of each plasmid copy falls HH-DNA + dNTP - ZHL-DNA, in inverse proportion to the concentration of the plasmid. There is no period with a total HGDNA + dNTP - HLDNA + LGDNA, switch-off of replication. In shifts from a low LGDNA + dNTP - ZLL-DNA. to a higher copy number, the replication On the assumption that the rate of each reprobability remains unchanged, whereas the action is proportional to the concentration of replication probability of each plasmid inthe DNA substrate, and that there is no difcreasesin inverse proportion to the concentration of the plasmid. There is no period with ‘Abbreviations used: HH, heavy DNA, HL, hybrid a burst of increased plasmid replication. DNA, LL, light DNA, ccc-DNA, covalently closed cirTherefore, plasmid RI does not fit into the cular DNA.
220
SHORT COMMUNICATIONS
ference in rate constant of the three reactions, the distribution between the three species (HH-, HL-, and LL-DNA) was calculated (curve A). The result did not fit the experimental data (circles). However, when a delay time was introduced such that the replication pattern did not reflect the composition of the plasmid pool at the replication time but at an earlier time (-td), it was possible to fit the experimental data using a tD = 0.22~ (7 = generation time) (curve B). This was concluded to be the eclipse period mentioned above (II). However, there is another, even more likely interpretation of the nature of td. If we assume that the newly replicated plasmid copies are not available for replication, becausethey are in a form that cannot be used for replication (Fig. 2), the mathematical predictions in Ref. (II) are fulfilled. The kinetics of the appearance of LGplasmid DNA will follow curve B assuming an average time of 0.22 generation lines, during which the replicating molecule and its daughter molecules are unable to participate in a new replication. Figure 2 was prepared to visualize the idea that replicating molecules and their daughters cannot participate in a second replication during a time interval, td: The figure demonstrates schematically the average replication pattern in an average cell. The number Density
of replication has been given the value of five and td is assumed to be 0.27. The five consecutive replication events are evenly spread over the whole cell cycle. Hence, the number of plasmid molecules that are unavailable for replication is always two in the average cell. The assumption that there is a considerable time during which plasmid molecules are unavailable for replication for a fairly long time is supported by experiments with so-called runaway replication derivatives of plasmid R 1. Although these retain at least some replication control activity (coded for by the copA gene (21)), the plasmid population grows exponentially with a generation time that is 0.4 of that of the host bacteria (22). Why newly replicated plasmid molecules are outside the replication pool is unknown. However, it may be appropriate to mention that the substrate for initiation of replication is supercoiled covalently closed circular (ccc) DNA (23). After replication, the plasmid molecules are present as ccc-DNA with a low superhelical density (Uhlin and Nordstrom, unpublished). The cycling time during runaway replication is prolonged by sublethal concentrations of the DNA gyrase inhibitor novobiocin (Uhlin and Nordstrom, unpublished). However, on the other hand, the cycling time is independent of the size of the molecule (23). There is also at least a second possibility to
5hift
Cell oiision
-li----~~~~~~~~ I
LL Ratia
LL/HL
0.00 -0.2
I 1 0
0.29 0.15 I 0.00
0.00 0.2
I ,
0.06
0.4 Time after
0.46 1 ,
0.6 density
o.Tl shift
0.66 I ,
0.29
0.11 1.0 Igmemtionsl
FIG. 2. Replication pattern during a density shift experiment with an exponentially growing population of plasmid-carrying bacteria. The 6gure shows an example of an average cell that contains five plasmid copies at birth and ten at cell division. The cell is in the middle of the cell cycle at zero time. Newly replicated molecules cannot take part in replication until 0.2 generations after initiation of replication. The other plasmid copies are selected randomly for replication. The values given in the figure are the average number of plasmid copies of each type (HH, HL, and LL) in or outside the replication pool in this class of cells.
221
SHORT COMMUNICATIONS
explain why newly replicated plasmid molecules are unavailable for replication. Replication of plasmid RI seemsto be limited by the availability of a protein coded for by the repA gene (24). There is some evidence that this protein is essentially cis acting (Molin and Light, unpublished) and that it is consumed during replication. Therefore, it may take some time to build up replication capacity before replication of a newly replicated plasmid molecule is possible. The picture that now emerges deviates somewhat from random selection. First, all plasmid copies in a cell are not always available for replication (they can be outside the replication pool). Secondly, there is a possibility that two or more plasmid replications coincide in time and that more than two plasmid copies are outside the replication pool. However, selection for replication is still basically random. The replication control system that now emergeswill be treated elsewhere. It predicts that irrespective of the copy number at birth, on the average, it copies will be replicated per cell and cell cycle. However, the system predicts a spread (presumably a Poissonian one) around the averagevalue n. One consequence of this is that the curve for LL-DNA/HL-DNA will be independent of the n value; this is not the case for the classical random selection model (15). In conclusion, there does not seem to be any contradiction between the results of the Meselson-Stahl experiments and the copy number shift experiments performed with plasmid R 1. ACKNOWLEDGMENT This work was supported by the Danish Medical Research Council (Project No. 227 I).
REFERENCES 1. KOLTER, R., AND HELINSKI, D. R., Annu. Rev. Genet. 13, 355-392 (1979). 2. PRITCHARD, R. H., In “Bacterial Episomes and Plasmids” (G. E. W. Wolstenholme and M. O’Connor, eds.), pp. 65-74, Churchill, London, 1969. PRITCHARD, R. H., BARTH, P. T., AND COLLINS, J.,
Symp. Sot. Gen. Microbial. 19, 263-297 (1969). MATXJBARA, K., ANDTAKEDA, Y., Mol. Gen. Genet. 142,225-230 (1975). MUROTSU, T., ANDMATSUBARA, K., Mol. Gen. Genet. 179, 509-519 (1980). MOLIN, S., STOUGAARD, P., LIGHT, J., NORDSTROM, M., ANDNORDSTROM, K., Mol. Gen. Genet. 181,
123-130 (1981). 7. ITOH, T., ANDTOMIZAWA, J.-I., Proc. Nat. Acad. Sci. USA 77,2450-2454 (1980). 8. MUESING, M., TAMM, J., SHEPARD, H. M., ANDPOLISKY, B., Cell 24, 235-242 (198 1).
9. TOMIZAWA, J.-I., ANDITOH, T., Proc. Nat. Acad. Sci. USA 78,6096-6100 (1981). 10. STOUGAARD, P., MOLIN, S., AND NORDSTROM, K., Proc. Nat. Acad. Sci. USA 78,6008-60 12 (198 1). 11. GUSTAFSSON, P., NORDSTROM, K., AND PERRAM, J. W., Plasmid 1, 187-203 (1978). 12. STEINBERG, D. A., ANDHELMSTETTER, C. A., Plasmid 6, 342-353 (198 1). 13. MESELSON, M., ANDSTAHL, F. W., Proc. Nat. Acad. Sci. USA 44,671-682 (1978). 14. BAZARAL, M., AND HELINSKI, D. R., Biochemistry 9, 399-406 (1970). 15. ROWND, R., J. Mol. Biol. 44, 387-402 (1969). 16. GUSTAFS~~N, P., AND NORDSTROM, K., J. Bacterial. 123,443-448
(1973).
17. UHLIN, B. E., AND NORDSTROM, K., J. Eacteriol. 124, 641-649 (1975). 18. MOLIN, S., AND NORDSTROM, K., J. BacterioL 141, 11l-120 (1980). 19. ENGBERG, B., AND NORDSTROM, K., J. Bacterial. 123, 179-186 (1975).
20. GUSTAFSSON,P., AND NORDSTROM, K., J. Bacterial. 141, 106-l 10 (1980).
21. MOLIN, S., DIAZ, R., UHLIN, B. E., AND NORDSTROM, K., J. Bacterial. 143, 1046-1048 (1980). 22. UHLIN, B. E., ANDNORDSTROM, K., Mol. Gen. Genet. 165, 167-179 (1978). 23. DIAZ, R., NORDSTROM, K., AND STAUDENBAUER, W. L., Nature (London) 289, 326-328 (198 I). 24. LIGHT, J., AND MOLIN, S., Mol. Gen. Genet. 184,5661 (1981).
9, 218-221 (1983)
Replication of Plasmid RI : Meselson-Stahl Density Shift Experiments Revisited KURT NORDSTROM’ Department of Molecular Biology, Odense University, Campusvej 55, DK-5230 Odense, Denmark Received April 5, 1982; revised November
16, 1982
Meselson-Stahl density shift experiments have been used extensively to study selection and timing of plasmid replication. Experiments with plasmid RI were previously performed and the conclusion was that this plasmid replicates one copy at a time and that there is an eclipse period after each replication during which no further replications can take place in the cell (Nordstrom et al., Plasmid 1, 187-203 (1978)). However, this interpretation is in conflict with other data, mainly with those obtained in copy number shift experiments (Gustafsson and Nordstrom, J. Bacterial. 141, 106-I 10 (1980)). However, the density shift experiments have now been reinterpreted such that there no longer is any conflict with the copy number shift experiments. There does not seem to be any such eclipse period, but newly replicated plasmid molecules are not available for a second replication for about 20% of a generation time.
Many plasmids are present in just a few (two-five) copies per bacterial cell. Nevertheless, such plasmids are very rarely lost. This demands the existance of replication control mechanisms, a prediction that has proven correct for a large number of plasmids. The plasmids themselves carry genes for control of their own replication (I). One property of a replication control system is that it should be able to correct deviations from the normal copy number. Therefore, a constitutively acting system analogous to a promoter system would not do. Pritchard (2,3) proposed his so-called inhibitor-dilution model, which among other things fulfils this correction requirement. The key element is an inhibitor that is formed in proportion to the concentration of its structural gene, i.e., in proportion to the concentration of the plasmid. Another key assumption of the model is that small fluctuations in inhibitor concentration result in large changes in replication probability. Replication inhibitors have been found for several plasmids. In some cases,the evidence is purely genetic (I); in others, the ’ Present addre~ Department of Microbiology, University of Uppsala, Biomedical Center, P.O. Box 581, S 751 23 Uppsala, Sweden. 0147-619X/83
$3.00
Copyright 0 1983 by Academic Press. Inc. All rights of reproduction in any form resews.
218
inhibitors have been identified as proteins Xdv (4.5) and R 1 (6)) or small RNA molecules (ColEl (7-9) and Rl (10)). Replication of plasmids seemsto be spread over the whole cell cycle (II) (with the possible exception of plasmid F, where the data are conflicting) (I I, 12). Furthermore, plasmid copies are selected randomly for replication; this has been demonstrated for many plasmids by Meselson-Stahl (13) density shift experiments (1 I, 14-16). Such experiments with plasmid Rl have been reported before (II). The following was concluded: (i) plasmid R 1 replicates one copy at a time; (iz) after each replication, there is an eclipse period during which no further replications can occur in the cell; (iii) the newly replicated molecules have the same probability as the other ones in the cell to be selected for the next replication (random selection); and (iv) plasmid replication is spread over the whole cell cycle. However, an increasing body of data obtained with FII plasmids (to which group Rl belongs) is in contradiction to the inhibitor dilution model. Plasmid R 100 (closely related to Rl) replicates with a significant rate even in the presenceof copy mutants of plasmid R 1, which indicates a sloppy responseto high concentrations of inhibitor (I 7). Cloned
SHORT COMMUNICATIONS I
rl
219
inhibitor dilution model (2,3), which predicts a period with total switch-off of plasmid replication in shifts from a high to a lower copy number and a burst of replication in the op posite shift. However, if the data contradict the idea that fluctuations in inhibitor concentration cause drastic changes in replication probability, an eclipse period after each replication ,' I OK&, cannot be explained by fluctuation in inhib1 0.5 0 itor concentration. It may even be possible Generatmns after density shkft that the whole idea of an eclipse period is FIG. 1. Appearance of LLplasmid DNA in an expowrong. Therefore, it may be appropriate to nentially growing population of bacteria during a density shift from heavy (H) to light (L) glycerol-minimal me- have a second look at the interpretation of experiment with respect dium (from Gustafsson et al. (II)). In curve A, the rate the Me&on-Stahl of formation of LLDNA and HLDNA is supposed to to plasmid replication. be=solely a function of the concentrations of LLDNA, In a density shift experiment, plasmid-carHL-DNA, and HH-DNA, which gives a relative amount tying bacteria are grown exponentially in a of HL-DNA = 2(eekL- eezk’)and of LL-DNA = I - 2eSk’ dense medium, e.g., “NH.++, D20, and then + e-*4 where k is the growth constant and t is time. In curve B, in addition to the assumptions made for curve shifted to light medium, e.g., 14NH4+,H20. A, there is a delay time (t,,)of 0.22 generation times, which The bacteria are labeled with [i4C]thymidine makes the kinetics dependent upon the previous history before and with [3H]thymidine after the shift. of the system. The experimentally found values for plasThe plasmid DNA is separated from chromid R I in Escherichia coli K- 12 strain EC 1005have been mosomal DNA and then analyzed for heavy inserted as open circles. (HH)-,* hybrid (HL)-, and light (LL)-DNA by fragments of the inhibitor gene do not com- CsCl density gradient centrif&ation (J I, Z4pletely switch off plasmid R 1 replication (18). 16). The appearance of LGDNA indicates Most important, though, is the result of copy plasmids that have replicated at least twice. For all plasmids studied, the result is banumber shift experiments. The copy number sically the same (plasmid F may be an excepof plasmid RI varies with the growth rate of tion (Rownd, personal communication)). Afthe host bacteria (19). Hence, it is possible to ter one generation, about % of the DNA is perform shift between different copy numbers HH, 3 HL, and ‘/5 LL, indicating random seby shifting the growth conditions (20). The lection for replication. The result of a density result from these experiments was that, irreshift experiment with plasmid RI is shown in spective of the copy number, on the average Fig. 1 (I 1). LGDNA started to appear after n copies are replicated per cell and cell cycle. about 0.3 generations. The data were interHence, in shifts from a high copy number to preted in the following way. There are three a lower one, the replication probability per reactions: cell remains unchanged, whereas the replication probability of each plasmid copy falls HH-DNA + dNTP - ZHL-DNA, in inverse proportion to the concentration of the plasmid. There is no period with a total HGDNA + dNTP - HLDNA + LGDNA, switch-off of replication. In shifts from a low LGDNA + dNTP - ZLL-DNA. to a higher copy number, the replication On the assumption that the rate of each reprobability remains unchanged, whereas the action is proportional to the concentration of replication probability of each plasmid inthe DNA substrate, and that there is no difcreasesin inverse proportion to the concentration of the plasmid. There is no period with ‘Abbreviations used: HH, heavy DNA, HL, hybrid a burst of increased plasmid replication. DNA, LL, light DNA, ccc-DNA, covalently closed cirTherefore, plasmid RI does not fit into the cular DNA.
220
SHORT COMMUNICATIONS
ference in rate constant of the three reactions, the distribution between the three species (HH-, HL-, and LL-DNA) was calculated (curve A). The result did not fit the experimental data (circles). However, when a delay time was introduced such that the replication pattern did not reflect the composition of the plasmid pool at the replication time but at an earlier time (-td), it was possible to fit the experimental data using a tD = 0.22~ (7 = generation time) (curve B). This was concluded to be the eclipse period mentioned above (II). However, there is another, even more likely interpretation of the nature of td. If we assume that the newly replicated plasmid copies are not available for replication, becausethey are in a form that cannot be used for replication (Fig. 2), the mathematical predictions in Ref. (II) are fulfilled. The kinetics of the appearance of LGplasmid DNA will follow curve B assuming an average time of 0.22 generation lines, during which the replicating molecule and its daughter molecules are unable to participate in a new replication. Figure 2 was prepared to visualize the idea that replicating molecules and their daughters cannot participate in a second replication during a time interval, td: The figure demonstrates schematically the average replication pattern in an average cell. The number Density
of replication has been given the value of five and td is assumed to be 0.27. The five consecutive replication events are evenly spread over the whole cell cycle. Hence, the number of plasmid molecules that are unavailable for replication is always two in the average cell. The assumption that there is a considerable time during which plasmid molecules are unavailable for replication for a fairly long time is supported by experiments with so-called runaway replication derivatives of plasmid R 1. Although these retain at least some replication control activity (coded for by the copA gene (21)), the plasmid population grows exponentially with a generation time that is 0.4 of that of the host bacteria (22). Why newly replicated plasmid molecules are outside the replication pool is unknown. However, it may be appropriate to mention that the substrate for initiation of replication is supercoiled covalently closed circular (ccc) DNA (23). After replication, the plasmid molecules are present as ccc-DNA with a low superhelical density (Uhlin and Nordstrom, unpublished). The cycling time during runaway replication is prolonged by sublethal concentrations of the DNA gyrase inhibitor novobiocin (Uhlin and Nordstrom, unpublished). However, on the other hand, the cycling time is independent of the size of the molecule (23). There is also at least a second possibility to
5hift
Cell oiision
-li----~~~~~~~~ I
LL Ratia
LL/HL
0.00 -0.2
I 1 0
0.29 0.15 I 0.00
0.00 0.2
I ,
0.06
0.4 Time after
0.46 1 ,
0.6 density
o.Tl shift
0.66 I ,
0.29
0.11 1.0 Igmemtionsl
FIG. 2. Replication pattern during a density shift experiment with an exponentially growing population of plasmid-carrying bacteria. The 6gure shows an example of an average cell that contains five plasmid copies at birth and ten at cell division. The cell is in the middle of the cell cycle at zero time. Newly replicated molecules cannot take part in replication until 0.2 generations after initiation of replication. The other plasmid copies are selected randomly for replication. The values given in the figure are the average number of plasmid copies of each type (HH, HL, and LL) in or outside the replication pool in this class of cells.
221
SHORT COMMUNICATIONS
explain why newly replicated plasmid molecules are unavailable for replication. Replication of plasmid RI seemsto be limited by the availability of a protein coded for by the repA gene (24). There is some evidence that this protein is essentially cis acting (Molin and Light, unpublished) and that it is consumed during replication. Therefore, it may take some time to build up replication capacity before replication of a newly replicated plasmid molecule is possible. The picture that now emerges deviates somewhat from random selection. First, all plasmid copies in a cell are not always available for replication (they can be outside the replication pool). Secondly, there is a possibility that two or more plasmid replications coincide in time and that more than two plasmid copies are outside the replication pool. However, selection for replication is still basically random. The replication control system that now emergeswill be treated elsewhere. It predicts that irrespective of the copy number at birth, on the average, it copies will be replicated per cell and cell cycle. However, the system predicts a spread (presumably a Poissonian one) around the averagevalue n. One consequence of this is that the curve for LL-DNA/HL-DNA will be independent of the n value; this is not the case for the classical random selection model (15). In conclusion, there does not seem to be any contradiction between the results of the Meselson-Stahl experiments and the copy number shift experiments performed with plasmid R 1. ACKNOWLEDGMENT This work was supported by the Danish Medical Research Council (Project No. 227 I).
REFERENCES 1. KOLTER, R., AND HELINSKI, D. R., Annu. Rev. Genet. 13, 355-392 (1979). 2. PRITCHARD, R. H., In “Bacterial Episomes and Plasmids” (G. E. W. Wolstenholme and M. O’Connor, eds.), pp. 65-74, Churchill, London, 1969. PRITCHARD, R. H., BARTH, P. T., AND COLLINS, J.,
Symp. Sot. Gen. Microbial. 19, 263-297 (1969). MATXJBARA, K., ANDTAKEDA, Y., Mol. Gen. Genet. 142,225-230 (1975). MUROTSU, T., ANDMATSUBARA, K., Mol. Gen. Genet. 179, 509-519 (1980). MOLIN, S., STOUGAARD, P., LIGHT, J., NORDSTROM, M., ANDNORDSTROM, K., Mol. Gen. Genet. 181,
123-130 (1981). 7. ITOH, T., ANDTOMIZAWA, J.-I., Proc. Nat. Acad. Sci. USA 77,2450-2454 (1980). 8. MUESING, M., TAMM, J., SHEPARD, H. M., ANDPOLISKY, B., Cell 24, 235-242 (198 1).
9. TOMIZAWA, J.-I., ANDITOH, T., Proc. Nat. Acad. Sci. USA 78,6096-6100 (1981). 10. STOUGAARD, P., MOLIN, S., AND NORDSTROM, K., Proc. Nat. Acad. Sci. USA 78,6008-60 12 (198 1). 11. GUSTAFSSON, P., NORDSTROM, K., AND PERRAM, J. W., Plasmid 1, 187-203 (1978). 12. STEINBERG, D. A., ANDHELMSTETTER, C. A., Plasmid 6, 342-353 (198 1). 13. MESELSON, M., ANDSTAHL, F. W., Proc. Nat. Acad. Sci. USA 44,671-682 (1978). 14. BAZARAL, M., AND HELINSKI, D. R., Biochemistry 9, 399-406 (1970). 15. ROWND, R., J. Mol. Biol. 44, 387-402 (1969). 16. GUSTAFS~~N, P., AND NORDSTROM, K., J. Bacterial. 123,443-448
(1973).
17. UHLIN, B. E., AND NORDSTROM, K., J. Eacteriol. 124, 641-649 (1975). 18. MOLIN, S., AND NORDSTROM, K., J. BacterioL 141, 11l-120 (1980). 19. ENGBERG, B., AND NORDSTROM, K., J. Bacterial. 123, 179-186 (1975).
20. GUSTAFSSON,P., AND NORDSTROM, K., J. Bacterial. 141, 106-l 10 (1980).
21. MOLIN, S., DIAZ, R., UHLIN, B. E., AND NORDSTROM, K., J. Bacterial. 143, 1046-1048 (1980). 22. UHLIN, B. E., ANDNORDSTROM, K., Mol. Gen. Genet. 165, 167-179 (1978). 23. DIAZ, R., NORDSTROM, K., AND STAUDENBAUER, W. L., Nature (London) 289, 326-328 (198 I). 24. LIGHT, J., AND MOLIN, S., Mol. Gen. Genet. 184,5661 (1981).