Vol. 134, No. 1
JOURNAL OF BACTERIOLOGY, Apr. 1978, p. 60-65 0021-9193/78/0134-0060$02.00/0 Copyright © 1978 American Society for Microbiology
Printed in U.S.A.
Probabilistic Behavior of DNA Segregation in Escherichia coli STEPHEN COOPER,* MARK SCHWIMMER, AND SARAH SCANLON Department ofMicrobiology, University of Michigan School of Medicine, Ann Arbor, Michigan 48109 Received for publication 24 December 1977
The pattern of segregation of DNA in Escherichia coli B/rK was analyzed by using the Methocel technique for forming chains of cells and the membrane binding elution method. Strain B/rK was shown to have a relatively high degree of nonrandom segregation and was used in a critical experiment to test the proposal that only one DNA strand acts nonrandomly during segregation. Thymidine-labeled cells were bound to a nitrocellulose membrane, and newly dividing cells were eluted from the membrane for six generations. The segregation of DNA in the eluted cells as well as in the cells bound to the membrane was examined by the Methocel technique. No difference in segregation was found between the two populations of cells, a result which indicates that the two strands are equivalent in segregation and that the pattern of segregation is not the result of a permanent bindinig of any strand to a pole of a cell.
studied strain [1]) that allows a clear distinction to be made between the two types of models.
DNA segregation in Escherichia coli has been shown to be nonrandom (1, 6, 7), although there had been previous suggestions to the contrary (4, 8). Although segregation is nonrandom (we define nonrandom to mean that there is a relationship between the pole toward which a DNA strand segregates in one generation and the pole that it segregates to in the succeeding generation), the precise pattern and mechanism of segregation remains an unsettled question. Two fundamentally different segregation models have been proposed. One type postulates that there is a permanent association of DNA to a pole of the cell. The Pierucci-Zuchowski (PZ) model is an example of this type, proposing that one strand of DNA segregates nonrandomly, as though it were bound to a pole of the cell, whereas the other, complementary strand, segregates randomly (7). A second type of model proposes that there is no distinction between strands and that any association between DNA and a pole is not permanent. The strand inertia (SI) model of Cooper and Weinberger (1) typifies this type of model. The SI model proposes that both strands can segregate nonrandomly and that any variability in the degree of nonrandomness (as observed in different media or different temperatures) can be accounted for by variation in the amount or strength of association between a strand of DNA and a pole of a cell. We now report a critical test to distinguish between these two models. The test has been made possible by the finding that E. coli B/rK has a relatively high degree of nonrandom segregation (as compared with B/rA, a previously
MATERIALS AND METHODS Segregation analysis was performed by using the Methocel technique of Lin et al. (4). The details of Methocel medium preparation, growth of cells, production of chains, fixation of slides, autoradiography, and data analysis have been described by Cooper and Weinberger (1). Cells were grown in MOPS medium (5) supplemented by the indicated carbon sources. The membrane binding and elution technique has been described previously (6). The particular characteristics of strains B/rK and B/rA have been described by Helmstetter and Pierucci (2). In all experiments described here, the segregation of DNA was observed in chains with one labeled cell. Since labeling a growing cell with thymidine produces at least two labeled DNA strands, the labeled cells were always allowed to grow in liquid (or on a membrane) for two to three generations before chain formation so that cells with one labeled strand of DNA were produced. It has been shown that after two generations of growth the segregation pattern reaches a steady state, which does not change with further growth (1).
RESULTS Segregation of DNA in E. coli B/rK. The relationship of DNA strand segregation to the poles of rod-shaped E. coli cells can be studied by allowing cells to form chains in viscous Methocel medium so that the orientation of cells produced by division is preserved. Pierucci and Zuchowski (7) analyzed thymidine-pulse-labeled cells that formed chains with two labeled cells. 60
VOL. 134, 1978
DNA SEGREGATION IN E. COLI
A simplification of this approach was introduced by Cooper and Weinberger (1), who grew labeled cells for a number of generations in liquid before chain formation, thus producing chains with only one labeled cell. The data can be presented in terms of the segregation ratio, R, which is the number of four-cell chains labeled in position 1 (pole position) divided by the number of fourcell chains labeled in position 2 (interior position). Eight cell chains can also yield the same data by considering that positions 1 and 2 and positions 3 and 4 of these chains were derived from positions 1 and 2, respectively, of four-cell chains. (Cell numbering in a chain always starts from the pole closest to a labeled cell.) The data from four- and eight-cell chains can be added to give a pooled ratio, represented as Rf/2 (1). The segregation ratios of E. coli B/rK in different media are given in Table 1. The ratios are compared with those obtained for similar experiments with B/rA (1). At all growth rates the degree of nonrandom segregation in B/rK is greater than that in B/rA, and the segregation ratio increases to a value greater than 3.0 at the slowest growth rates. Two observations can be made from these data, as follows. (i) The fact that segregation ratios greater than 3.0 are not predicted by the PZ model (7) may be taken as preliminary evidence against the model. [This prediction can be understood by considering a collection of cells, each with one labeled strand, half + and half -. If chains are prepared from these cells, then according to the PZ model, all of the + strands will be found in position 1 of the four-cell chain, and the strains will be found equally in positions 1 and 2. Therefore, the ratio of total chains found in positions 1 and 2 is (½ + Y):(Y) or 3:1.] This type of result indicates that there must be some nonrandom segregation behavior in both
61
strands, although it does not eliminate the possibility of one strand acting completely or almost completely nonrandomly while the other strand exhibits only a small amount of nonrandom segregation behavior. Statistical analysis (chi square) indicates that the experimental R value for glycerol is significantly different from the predicted 3.0 of the PZ model. This segregation value of greater than 3.0 may not be sufficient, by itself, to disprove the PZ model. It is possible that difficulties in chain counting and errors arising from human bias can account for the differences between the predicted and experimental R values. These possible errors were minimized by counting at least 1,000 chains for each experimental condition and having the counting done by a number of people. (ii) A strain with a high degree of nonrandom segregation is a suitable candidate for analysis of segregation in a critical experiment to test the PZ model as described below. Cells with more random segregation ratios (as found in B/rA [1]) cannot be used to differentiate different models, and, therefore, the previously studied strain, B/rA, was not suitable for a critical test of the different models (Table 2). Membrane elution experiment. (i) Rationale of the ewperiment. The PZ model of segregation predicts that it is possible to distinguish the two different strands of DNA that segregate in either a random or nonrandom manner, whereas the SI model predicts that no such distinction between the two strands exists. The membrane elution technique allows a testing of these predictions (Fig. 1). The PZ model predicts that labeled strands permanently bound to the pole of a cell will be preferentially found in the fraction of cells bound to the Millipore membrane, whereas this is not predicted by the SI model. (It should be noted that the use of the
TABLE 1. Comparison of DNA segregation in E. coli B/rK and B/rA grown in glycerol, glucose, and glucose-Casamino Acids at 370Cm Medium
Doubling (h)
No. of chains counted with radioactivity observed in mdicated position of cell chain eight-cell chain four-cell chain
B/rK 1
Glycerol
2
1
2
3
B/rA (1)
4
481 314 136 54 1.05 33 63 3.66 1.85 1.33 930 Glucose 316 209 66 38 44 3.03 1.56 o 1.76 390 Glucose-C 202 246 163 116 100 1.91 1.16 Acids aE. coli B/rK was grown in the indicated medium, labeled for 4 min with [3H]thymidine, washed by centrifugation, and resuspended in fresh medium. The cells were grown for at least two generations, collected by centrifugation, suspended in Methocel medium containing the same carbon source as the cells had been grown on, and spread on slides for chain formation. After autoradiography, the labeled cells in the indicated positions were noted and summaized. RI/2 - sum of position 1 of four-cell chains and position 1 and 2 of eightcell chains divided by the sum of position 2 of four-cell chains and positions 3 and 4 of eight-cell chains. Values for strain B/rA are from Cooper and Weinberger (1).
62
COOPER, SCHWIMMER, AND SCANLON
J. BACTERIOL.
TABLE 2. Relative numbers of labeled + and - strains eluted from filter-bound cells according to different segregation models Generation of elution
-eluted Strand
+eluted Strand
- maining Strand re-
+ maing Strand re-
Rl' eluate
Rf/2brane mem-
3.0 2.3 5.0
3.0 5.0 5.0 9.0 17.0 33.0
P-Z model (1) 1 2a 3 4 5 6
100 75 12.5 6.25 3.13 1.56
100 50 25 0 0 0
100 25 12.5 6.25 3.13 1.56
100 50 25 25 25 25
1.0 1.0 1.0
SI model (P = 0.75) (2)
100 75 6.25 4.69 3.52 2.64
100 100 100 3.0 3.0 75 25 25 3.0 3.0 3 6.25 18.75 18.75 3.0 3.0 4 4.69 14.06 14.06 3.0 3.0 5 3.52 10.56 10.56 3.0 3.0 6 2.64 7.92 7.92 3.0 3.0 a The apparently paradoxical preferential elution of DNA in the PZ model is fully explained by Pierucci and Helmstetter (6). Similar considerations apply to the SI model as noted in the text. Cells with segregation ratios much less than 3.0 yield membrane fractions with significant decreased R values such that a clear distinction between the two models is precluded. This is the reason that strain B/rA was unsuitable for the critical experiment (Table 1). In the SI model, P is the probability that a strand will segregate toward the same pole that it segregated to in the previous generation (1). 1
2b
word "binding" does not necessarily imply any physical mode of attachment of DNA to the cell pole, but rather the formal notion of apparent binding.) Figure 1 shows that the R value increases in the membrane-bound cells according to the PZ model (Fig. 1A), but does not increase according to the SI model (Fig. 1B). The two different models, or classes of models, lead to two different predictions regarding the segregation ratios of the cells in the eluate and of the cells remaining bound to the membrane after a number of generations of growth on the membrane. The expected R values are given more precisely in Table 2, where the PZ model is contrasted with the predictions of the SI model for a segregation ratio of 3.0. (The SI model can vary to account for any R value, but the best test of the PZ model comes from a cell that has the R value close to that predicted by the PZ model, 3.0.) (ii) Segregation of DNA in cells eluted from and bound to a membrane. E. coli B/rK grown in glucose-MOPS at 37°C was labeled for 4 min with [3H]thymidine and filtered onto a Millipore membrane. The membrane was inverted, and fresh medium was pumped through it to elute the dividing cells (Fig. 2). To demonstrate that this experimental procedure was valid, the number of cells bound and eluted at various times was monitored, as was the radioactivity per cell. These measurements indicated that: (i) the number of cells bound to the membrane (as originally determined by the difference between the number of cells filtered onto the
membrane and the number eluted off in the first few minutes of elution) was approximately equal to the number of cells found on the membrane at the end of the experiment and to the number of cells eluted in any one generation; (ii) the patterns of plateaus and decreases in radioactivity per cell indicate a growth rate on the membrane that is the same as the growth rate in liquid; and (iii) the decreases in radioactivity at C + D minutes and C + D + r minutes (Fig. 2; see Pierucci and Helmstetter for detailed discussion [3]) were less than and greater than a factor of two, respectively. These three results indicate that the cells bound to the membrane are growing normally, with no significant proportion of nongrowing cells or slow-growing cells, and that there is some asymmetry in the elution of labeled DNA that should give the proper separation of + and - strands if the PZ model is correct. At various times during elution, samples of the eluate were taken and either grown further in liquid or processed immediately for chain formation (Fig. 2). In all cases, cells were allowed to grow for at least three generations either in liquid or on the membrane before placement in Methocel. The R values determined for each of these fractions, as well as those determined for the unfractionated cells originally not bound to the membrane (wash-off), are given in Table 3. There is absolutely no indication that the R value decreases with time in the eluate, nor is there any indication of a high R value in the membrane-bound fraction as predicted by the
VOL. 134, 1978 A.
DNA SEGREGATION IN E. COLI
MEMBRANE
ELUATE
344()34~144 1022( G 0 0 o-u 542 1
101
(
r
)
94
B.
344() 344 311 (
333 1) 322
FIG. 1. Elution pattern of labeled strands according to the PZ and SI models or models of a similar type. (A) Pattern according to the PZ model in which + strands are permanently bound to a pole of a cell. The + and - indicate thepresence ofa labeled strand of one or the other complementarity. Complementary unlabeled strands are present but not indicated. Note that after the first generation no further elution of + strands takes place, and the - strands segregate randomly. The predicted R values obtained from the various mixtures of + and - strands in the eluate or on the membrane are given in boldface at the sides. For example, the R value in line 3 is derived by considering 10 labeled ceUs, 8 with labeled + strands and 2 with labeled - strands. The former cells produce four-ceU chains aU labeled in position 1, and the latter cells produce four-ceU chains with one labeled in position 1 and the other labeled in position 2. Therefore, the R value is predicted to be 9 (nine chains labeled in position I divided by one chain labeled in position 2). (B) IUustration of the type of pattern expected in the SI model in which a strand preferentialy segregates in the same direction as in the previous ceU division. There is no permanent attachment, and the two strands behave in the same manner. The actual numbers of + and - strands eluted are not meant to be a mathematically exact representation, but merely represent the idea that the strands preferentially remain attached to the membrane. Table 2 gives the precise values for the elution pattern according to the two models.
PZ model (Table 2). Similar results have been obtained for E. coli B/rK grown in glycerol-MOPS medium. In that experiment, the R values for the unselected cells (wash-off) and membrane-bound cells were 3.72 and 3.91, respectively (data not shown). Table 3 shows that there was no indication of a decreased eluate R value or any value for the membrane fraction approaching the expected R value of 33.0 (Table 2) as predicted by the PZ model. We therefore conclude that the data are not consistent with the PZ model for nonrandom
63
DNA segregation. The results are consistent with the SI model.
DISCUSSION PZ model. The data are not consistent with the prediction of the PZ model (7), that one strand will act nonrandomly as though bound to one pole of the cell and that the other strand will act randomly. There is no evidence of any distinction between strands of DNA and no evidence for a permanent attachment of DNA to a pole. The segregation ratios reported for E. coli B/rK during slow growth, as well as the variation in segregation patterns observed in different media, can also be taken as evidence against the PZ model. SI model. The SI model (1) is a probabilistic model for segregation, which does not postulate a difference between DNA strands. It suggests that there is a variable tendency for each strand to segregate toward the pole to which it segregated at the previous cell division. The data reported in this paper are consistent with this model. Further testing of this model is required, however, since our results do not specifically eliminate other probabilistic models for nonrandom segregation. There are two ways of stating the SI model. They differ only in the emphasis each puts upon the actual nonrandom segregation observed, and we have been unable to distinguish between the two experimentally. These two statements are: (i) segregation is an inherently random process, but superimposed on this pattern is some unessential nonrandom behavior, and (ii) segregation is a fundamentally nonrandom process, but superimposed upon this process is some randomizing event(s) that can reduce and vary the observed nonrandom segregation. It is difficult to foresee how one can distinguish the above two cases. At any rate, the distinction may not be of value in understanding DNA segregation. Probabilistic versus deterministic segregation. The PZ model is an example of a general class of models in which an irreversible attachment of DNA is made to one of the poles of a cell. The SI model is an example of a general class of models in which any "attachment" made is temporary and reversible and in which all strands segregate in a similar manner. The data in this paper suggest that all models of the type that postulates an irreversible attachment of the PZ type are unsatisfactory and that only models of the SI type can be entertained. Possible artifacts and limitations of the criticai experiment. It is possible that some unknown features of the Methocel technique (that may alter the pattern of segregation nor-
64
COOPER, SCHWIMMER, AND SCANLON
J. BACTERIOL.
EXPONENTIALLY GROWING CULTURE OF E. col B/r K F H3 THYMKINE 4 MNUTE INCUBATION
PUT ON MEMBRANE, ,IN. INVERT, ELUTE WITH FRESH MEDIUM
0
WASHOFF
I
C+D '
i
I C+D+ 7-
SLIDES
ELWATE
2
ELUATE
I SUDES
SLIDES
3
SLIDES
.4 l 4 CELL
SLIDES .ELSUATE
NINg (44Id)
S
CELL
4s 4Is
SLIDES
4's
MEMNE ELITE
CHAINS(Wo) SLIDES
I's
I
a'
5 6
SLIDES
4's 4's
I a's
4
as4'
a'
8's
7
*8 .9 TIME (GENERATIONS)
FIG. 2. Protocol of a membrane elution-Methocel segregation experiment. A total of 100 ml of a glucoseMOPS culture of E. coli B/rK was labeled for 4 min with 3H]thymidine (60.83 Ci/mmol; 10 ,Ci/ml) and filtered onto a Millipore membrane (type GS, 142-mm diameter). After the filter was washed with MOPSglucose medium, it was inverted and fresh medium was pumped through it (6). The first 4 min of elution yields the "wash-off," which is the ceUs not bound to the membrane. These are unselected cells and are grown for two to three generations before placement in Methocel so that cells with only one labeled strand are produced. Similarly, cells eluted after one or two generations of growth on the membrane are allowed to incubate in liquid for an additional period to assure the complete segregation of DNA strands. From the third generation of elution onward, there is no need for additional incubation in liquid culture as growth on the membrane is sufficient to allow the segregation of the DNA strands. After six generations of elution, the cells from the membrane are washed off with a rubber policeman and fresh medium, and the cells so obtained are placed on slides for chain formation. C + D and C + D + T are the times (6) at which decreases in radioactivity of less than and more than a factor of two, respectively, wiU occur according to both the PZ and SI models (1, 7).
mally present in liquid media) or the membrane binding technique (that may not release cells, as suggested in Fig. 1) may affect and weaken the critical experiment presented above. These possibilities appear unlikely for the following two reasons. First, our experiments are designed to detect differences and trends in segregation by using the Methocel technique throughout, thus minimizing the effect of any uniform error. It would be expected that, if the Methocel technique altered the segregation values in any particular sample of cells, one would still be able to observe any gross changes or trends predicted by the PZ model. Second, evidence that the membrane elution technique is working as expected is given by the variation in the elution
plateau ratios as described by Pierucci and Helmstetter (6) and as also observed by us in these experiments. First and second segregation. After completion of the synthesis of a new DNA strand, it will segregate into one of two daughter cells at the next cell division. What will happen if the pole toward which the new strand segregated is also preferentially chosen as the pole toward which the strand will segregate at the next cell division? Since the first segregation occurred with, presumably, an associated older strand that was the template for the new strand, at the second division this older strand must segregate away from the new strand and toward the pole to which it did not segregate at the first division.
DNA SEGREGATION IN E. COLI
VoL- 134, 1978
TABLE 3. Comparison of experimental and predicted segregation ratios for membrane elution experiment in glucose at 370C Calculated R value
Sample of cells
analyzed
Experimental PZ
PZmodel 3.0
SI model (P
R value
-0.75)a
3.0 2.84 Wash-off (unselected) 3.0 3.0 Generation 1 2.53 eluate 2.3 3.0 Generation 2 2.92 eluate 5.0 3.0 2.72 Generation 3 eluate 1.0 3.0 Generation 4 3.05 eluate 1.0 3.0 Generation 5 2.93 eluate 1.0 3.0 2.71 Generation 6 eluate 33.0 3.0 Membrane 3.07 a P is the probability that a strand will segregate toward the same pole that it segregated to in the previous generation.
This paradoxical situation means that all older strands will actually segregate preferentially toward the inner cells of chains rather than toward the outer cells of chains as observed. Thus, we must conclude that in any situation in which nonrandom segregation is taking place there must be some "seniority" system in which the older strands segregate preferentially toward the pole to which they segregated at the previous division. In the PZ model this is stated as the requirement that "attachment" to a pole does not occur until the newly synthesized strand is first used as a template. We merely wish to show that this type of rule (that attachment to a pole occurs when the strand is first used as a template) must be present in any system of nonrandom segregation of complementary strands, and therefore this type of strand movement is incor-
65
porated into the SI model (1). Mechanism of segregation. None of the results reported here bears directly on the mechanism of nonrandom segregation. However, if segregation is dependent on some DNA-membrane association, then the variable nonrandom segregation may be due to either the variation in the attachment of the DNA to the membrane or the variation of the association of the membrane to a particular pole of the bacterial cell
(1). ACKNOWLEDGMENT This work was supported by grant GB 40099 from the National Science Foundation.
LITERATURE CITED 1. Cooper, S., and M. Weinberger. 1977. Medium-dependent variation of deoxyribonucleic acid segregation in Escherichia coli. J. Bacteriol. 130:118-127. 2. Helmstetter, C. E., and 0. Pierucci. 1976. DNA synthesis during the division cycle of three substrains of Escherichia coli B/r. J. Mol. Biol. 102:477-486. 3. Lark, K. G., EL Eberle, R. A. Consgli, EL C. Minocha, N. Chai, and C. Lark. 1967. Chromosome segregation and the regulation of DNA replication, p. 63-89. In H. J. Vogel, J. 0. Lampen, and V. Bryson (ed.), Organizational biosynthesis. Academic Press Inc., New York. 4. Lin, E. C. C., Y. Hirota, and F. Jacob. 1971. On the process of cellular division in Escherichia coli. VI. Use of a Methocel-autoradiographic method for the study of cellular division in Escherichia coli. J. Bacteriol. 108:375-385. 5. Neidhardt, F. C., P. L. Bloch, and D. F. Smith. 1974. Culture medium for enterobacteria. J. Bacteriol. 119:736-747. 6. Pierucci, O., and C. E. Helmstetter. 1976. Chromosome segregation in Escherichia coli B/r at various growth rates. J. Bacteriol. 128:708-716. 7. Pierucci, O., and C. Zuchow8ki. 1973. Non-random segregation of DNA strands in Escherichia coli B/r. J. Mol. Biol. 80:477-503. 8. Ryter, A., Y. Hirota, and F. Jacob. 1968. DNA-membrane complex and nuolear segregation in bacteria. Cold Spring Harbor Symp. Quant. Biol. 33:669-676.
JOURNAL OF BACTERIOLOGY, Apr. 1978, p. 60-65 0021-9193/78/0134-0060$02.00/0 Copyright © 1978 American Society for Microbiology
Printed in U.S.A.
Probabilistic Behavior of DNA Segregation in Escherichia coli STEPHEN COOPER,* MARK SCHWIMMER, AND SARAH SCANLON Department ofMicrobiology, University of Michigan School of Medicine, Ann Arbor, Michigan 48109 Received for publication 24 December 1977
The pattern of segregation of DNA in Escherichia coli B/rK was analyzed by using the Methocel technique for forming chains of cells and the membrane binding elution method. Strain B/rK was shown to have a relatively high degree of nonrandom segregation and was used in a critical experiment to test the proposal that only one DNA strand acts nonrandomly during segregation. Thymidine-labeled cells were bound to a nitrocellulose membrane, and newly dividing cells were eluted from the membrane for six generations. The segregation of DNA in the eluted cells as well as in the cells bound to the membrane was examined by the Methocel technique. No difference in segregation was found between the two populations of cells, a result which indicates that the two strands are equivalent in segregation and that the pattern of segregation is not the result of a permanent bindinig of any strand to a pole of a cell.
studied strain [1]) that allows a clear distinction to be made between the two types of models.
DNA segregation in Escherichia coli has been shown to be nonrandom (1, 6, 7), although there had been previous suggestions to the contrary (4, 8). Although segregation is nonrandom (we define nonrandom to mean that there is a relationship between the pole toward which a DNA strand segregates in one generation and the pole that it segregates to in the succeeding generation), the precise pattern and mechanism of segregation remains an unsettled question. Two fundamentally different segregation models have been proposed. One type postulates that there is a permanent association of DNA to a pole of the cell. The Pierucci-Zuchowski (PZ) model is an example of this type, proposing that one strand of DNA segregates nonrandomly, as though it were bound to a pole of the cell, whereas the other, complementary strand, segregates randomly (7). A second type of model proposes that there is no distinction between strands and that any association between DNA and a pole is not permanent. The strand inertia (SI) model of Cooper and Weinberger (1) typifies this type of model. The SI model proposes that both strands can segregate nonrandomly and that any variability in the degree of nonrandomness (as observed in different media or different temperatures) can be accounted for by variation in the amount or strength of association between a strand of DNA and a pole of a cell. We now report a critical test to distinguish between these two models. The test has been made possible by the finding that E. coli B/rK has a relatively high degree of nonrandom segregation (as compared with B/rA, a previously
MATERIALS AND METHODS Segregation analysis was performed by using the Methocel technique of Lin et al. (4). The details of Methocel medium preparation, growth of cells, production of chains, fixation of slides, autoradiography, and data analysis have been described by Cooper and Weinberger (1). Cells were grown in MOPS medium (5) supplemented by the indicated carbon sources. The membrane binding and elution technique has been described previously (6). The particular characteristics of strains B/rK and B/rA have been described by Helmstetter and Pierucci (2). In all experiments described here, the segregation of DNA was observed in chains with one labeled cell. Since labeling a growing cell with thymidine produces at least two labeled DNA strands, the labeled cells were always allowed to grow in liquid (or on a membrane) for two to three generations before chain formation so that cells with one labeled strand of DNA were produced. It has been shown that after two generations of growth the segregation pattern reaches a steady state, which does not change with further growth (1).
RESULTS Segregation of DNA in E. coli B/rK. The relationship of DNA strand segregation to the poles of rod-shaped E. coli cells can be studied by allowing cells to form chains in viscous Methocel medium so that the orientation of cells produced by division is preserved. Pierucci and Zuchowski (7) analyzed thymidine-pulse-labeled cells that formed chains with two labeled cells. 60
VOL. 134, 1978
DNA SEGREGATION IN E. COLI
A simplification of this approach was introduced by Cooper and Weinberger (1), who grew labeled cells for a number of generations in liquid before chain formation, thus producing chains with only one labeled cell. The data can be presented in terms of the segregation ratio, R, which is the number of four-cell chains labeled in position 1 (pole position) divided by the number of fourcell chains labeled in position 2 (interior position). Eight cell chains can also yield the same data by considering that positions 1 and 2 and positions 3 and 4 of these chains were derived from positions 1 and 2, respectively, of four-cell chains. (Cell numbering in a chain always starts from the pole closest to a labeled cell.) The data from four- and eight-cell chains can be added to give a pooled ratio, represented as Rf/2 (1). The segregation ratios of E. coli B/rK in different media are given in Table 1. The ratios are compared with those obtained for similar experiments with B/rA (1). At all growth rates the degree of nonrandom segregation in B/rK is greater than that in B/rA, and the segregation ratio increases to a value greater than 3.0 at the slowest growth rates. Two observations can be made from these data, as follows. (i) The fact that segregation ratios greater than 3.0 are not predicted by the PZ model (7) may be taken as preliminary evidence against the model. [This prediction can be understood by considering a collection of cells, each with one labeled strand, half + and half -. If chains are prepared from these cells, then according to the PZ model, all of the + strands will be found in position 1 of the four-cell chain, and the strains will be found equally in positions 1 and 2. Therefore, the ratio of total chains found in positions 1 and 2 is (½ + Y):(Y) or 3:1.] This type of result indicates that there must be some nonrandom segregation behavior in both
61
strands, although it does not eliminate the possibility of one strand acting completely or almost completely nonrandomly while the other strand exhibits only a small amount of nonrandom segregation behavior. Statistical analysis (chi square) indicates that the experimental R value for glycerol is significantly different from the predicted 3.0 of the PZ model. This segregation value of greater than 3.0 may not be sufficient, by itself, to disprove the PZ model. It is possible that difficulties in chain counting and errors arising from human bias can account for the differences between the predicted and experimental R values. These possible errors were minimized by counting at least 1,000 chains for each experimental condition and having the counting done by a number of people. (ii) A strain with a high degree of nonrandom segregation is a suitable candidate for analysis of segregation in a critical experiment to test the PZ model as described below. Cells with more random segregation ratios (as found in B/rA [1]) cannot be used to differentiate different models, and, therefore, the previously studied strain, B/rA, was not suitable for a critical test of the different models (Table 2). Membrane elution experiment. (i) Rationale of the ewperiment. The PZ model of segregation predicts that it is possible to distinguish the two different strands of DNA that segregate in either a random or nonrandom manner, whereas the SI model predicts that no such distinction between the two strands exists. The membrane elution technique allows a testing of these predictions (Fig. 1). The PZ model predicts that labeled strands permanently bound to the pole of a cell will be preferentially found in the fraction of cells bound to the Millipore membrane, whereas this is not predicted by the SI model. (It should be noted that the use of the
TABLE 1. Comparison of DNA segregation in E. coli B/rK and B/rA grown in glycerol, glucose, and glucose-Casamino Acids at 370Cm Medium
Doubling (h)
No. of chains counted with radioactivity observed in mdicated position of cell chain eight-cell chain four-cell chain
B/rK 1
Glycerol
2
1
2
3
B/rA (1)
4
481 314 136 54 1.05 33 63 3.66 1.85 1.33 930 Glucose 316 209 66 38 44 3.03 1.56 o 1.76 390 Glucose-C 202 246 163 116 100 1.91 1.16 Acids aE. coli B/rK was grown in the indicated medium, labeled for 4 min with [3H]thymidine, washed by centrifugation, and resuspended in fresh medium. The cells were grown for at least two generations, collected by centrifugation, suspended in Methocel medium containing the same carbon source as the cells had been grown on, and spread on slides for chain formation. After autoradiography, the labeled cells in the indicated positions were noted and summaized. RI/2 - sum of position 1 of four-cell chains and position 1 and 2 of eightcell chains divided by the sum of position 2 of four-cell chains and positions 3 and 4 of eight-cell chains. Values for strain B/rA are from Cooper and Weinberger (1).
62
COOPER, SCHWIMMER, AND SCANLON
J. BACTERIOL.
TABLE 2. Relative numbers of labeled + and - strains eluted from filter-bound cells according to different segregation models Generation of elution
-eluted Strand
+eluted Strand
- maining Strand re-
+ maing Strand re-
Rl' eluate
Rf/2brane mem-
3.0 2.3 5.0
3.0 5.0 5.0 9.0 17.0 33.0
P-Z model (1) 1 2a 3 4 5 6
100 75 12.5 6.25 3.13 1.56
100 50 25 0 0 0
100 25 12.5 6.25 3.13 1.56
100 50 25 25 25 25
1.0 1.0 1.0
SI model (P = 0.75) (2)
100 75 6.25 4.69 3.52 2.64
100 100 100 3.0 3.0 75 25 25 3.0 3.0 3 6.25 18.75 18.75 3.0 3.0 4 4.69 14.06 14.06 3.0 3.0 5 3.52 10.56 10.56 3.0 3.0 6 2.64 7.92 7.92 3.0 3.0 a The apparently paradoxical preferential elution of DNA in the PZ model is fully explained by Pierucci and Helmstetter (6). Similar considerations apply to the SI model as noted in the text. Cells with segregation ratios much less than 3.0 yield membrane fractions with significant decreased R values such that a clear distinction between the two models is precluded. This is the reason that strain B/rA was unsuitable for the critical experiment (Table 1). In the SI model, P is the probability that a strand will segregate toward the same pole that it segregated to in the previous generation (1). 1
2b
word "binding" does not necessarily imply any physical mode of attachment of DNA to the cell pole, but rather the formal notion of apparent binding.) Figure 1 shows that the R value increases in the membrane-bound cells according to the PZ model (Fig. 1A), but does not increase according to the SI model (Fig. 1B). The two different models, or classes of models, lead to two different predictions regarding the segregation ratios of the cells in the eluate and of the cells remaining bound to the membrane after a number of generations of growth on the membrane. The expected R values are given more precisely in Table 2, where the PZ model is contrasted with the predictions of the SI model for a segregation ratio of 3.0. (The SI model can vary to account for any R value, but the best test of the PZ model comes from a cell that has the R value close to that predicted by the PZ model, 3.0.) (ii) Segregation of DNA in cells eluted from and bound to a membrane. E. coli B/rK grown in glucose-MOPS at 37°C was labeled for 4 min with [3H]thymidine and filtered onto a Millipore membrane. The membrane was inverted, and fresh medium was pumped through it to elute the dividing cells (Fig. 2). To demonstrate that this experimental procedure was valid, the number of cells bound and eluted at various times was monitored, as was the radioactivity per cell. These measurements indicated that: (i) the number of cells bound to the membrane (as originally determined by the difference between the number of cells filtered onto the
membrane and the number eluted off in the first few minutes of elution) was approximately equal to the number of cells found on the membrane at the end of the experiment and to the number of cells eluted in any one generation; (ii) the patterns of plateaus and decreases in radioactivity per cell indicate a growth rate on the membrane that is the same as the growth rate in liquid; and (iii) the decreases in radioactivity at C + D minutes and C + D + r minutes (Fig. 2; see Pierucci and Helmstetter for detailed discussion [3]) were less than and greater than a factor of two, respectively. These three results indicate that the cells bound to the membrane are growing normally, with no significant proportion of nongrowing cells or slow-growing cells, and that there is some asymmetry in the elution of labeled DNA that should give the proper separation of + and - strands if the PZ model is correct. At various times during elution, samples of the eluate were taken and either grown further in liquid or processed immediately for chain formation (Fig. 2). In all cases, cells were allowed to grow for at least three generations either in liquid or on the membrane before placement in Methocel. The R values determined for each of these fractions, as well as those determined for the unfractionated cells originally not bound to the membrane (wash-off), are given in Table 3. There is absolutely no indication that the R value decreases with time in the eluate, nor is there any indication of a high R value in the membrane-bound fraction as predicted by the
VOL. 134, 1978 A.
DNA SEGREGATION IN E. COLI
MEMBRANE
ELUATE
344()34~144 1022( G 0 0 o-u 542 1
101
(
r
)
94
B.
344() 344 311 (
333 1) 322
FIG. 1. Elution pattern of labeled strands according to the PZ and SI models or models of a similar type. (A) Pattern according to the PZ model in which + strands are permanently bound to a pole of a cell. The + and - indicate thepresence ofa labeled strand of one or the other complementarity. Complementary unlabeled strands are present but not indicated. Note that after the first generation no further elution of + strands takes place, and the - strands segregate randomly. The predicted R values obtained from the various mixtures of + and - strands in the eluate or on the membrane are given in boldface at the sides. For example, the R value in line 3 is derived by considering 10 labeled ceUs, 8 with labeled + strands and 2 with labeled - strands. The former cells produce four-ceU chains aU labeled in position 1, and the latter cells produce four-ceU chains with one labeled in position 1 and the other labeled in position 2. Therefore, the R value is predicted to be 9 (nine chains labeled in position I divided by one chain labeled in position 2). (B) IUustration of the type of pattern expected in the SI model in which a strand preferentialy segregates in the same direction as in the previous ceU division. There is no permanent attachment, and the two strands behave in the same manner. The actual numbers of + and - strands eluted are not meant to be a mathematically exact representation, but merely represent the idea that the strands preferentially remain attached to the membrane. Table 2 gives the precise values for the elution pattern according to the two models.
PZ model (Table 2). Similar results have been obtained for E. coli B/rK grown in glycerol-MOPS medium. In that experiment, the R values for the unselected cells (wash-off) and membrane-bound cells were 3.72 and 3.91, respectively (data not shown). Table 3 shows that there was no indication of a decreased eluate R value or any value for the membrane fraction approaching the expected R value of 33.0 (Table 2) as predicted by the PZ model. We therefore conclude that the data are not consistent with the PZ model for nonrandom
63
DNA segregation. The results are consistent with the SI model.
DISCUSSION PZ model. The data are not consistent with the prediction of the PZ model (7), that one strand will act nonrandomly as though bound to one pole of the cell and that the other strand will act randomly. There is no evidence of any distinction between strands of DNA and no evidence for a permanent attachment of DNA to a pole. The segregation ratios reported for E. coli B/rK during slow growth, as well as the variation in segregation patterns observed in different media, can also be taken as evidence against the PZ model. SI model. The SI model (1) is a probabilistic model for segregation, which does not postulate a difference between DNA strands. It suggests that there is a variable tendency for each strand to segregate toward the pole to which it segregated at the previous cell division. The data reported in this paper are consistent with this model. Further testing of this model is required, however, since our results do not specifically eliminate other probabilistic models for nonrandom segregation. There are two ways of stating the SI model. They differ only in the emphasis each puts upon the actual nonrandom segregation observed, and we have been unable to distinguish between the two experimentally. These two statements are: (i) segregation is an inherently random process, but superimposed on this pattern is some unessential nonrandom behavior, and (ii) segregation is a fundamentally nonrandom process, but superimposed upon this process is some randomizing event(s) that can reduce and vary the observed nonrandom segregation. It is difficult to foresee how one can distinguish the above two cases. At any rate, the distinction may not be of value in understanding DNA segregation. Probabilistic versus deterministic segregation. The PZ model is an example of a general class of models in which an irreversible attachment of DNA is made to one of the poles of a cell. The SI model is an example of a general class of models in which any "attachment" made is temporary and reversible and in which all strands segregate in a similar manner. The data in this paper suggest that all models of the type that postulates an irreversible attachment of the PZ type are unsatisfactory and that only models of the SI type can be entertained. Possible artifacts and limitations of the criticai experiment. It is possible that some unknown features of the Methocel technique (that may alter the pattern of segregation nor-
64
COOPER, SCHWIMMER, AND SCANLON
J. BACTERIOL.
EXPONENTIALLY GROWING CULTURE OF E. col B/r K F H3 THYMKINE 4 MNUTE INCUBATION
PUT ON MEMBRANE, ,IN. INVERT, ELUTE WITH FRESH MEDIUM
0
WASHOFF
I
C+D '
i
I C+D+ 7-
SLIDES
ELWATE
2
ELUATE
I SUDES
SLIDES
3
SLIDES
.4 l 4 CELL
SLIDES .ELSUATE
NINg (44Id)
S
CELL
4s 4Is
SLIDES
4's
MEMNE ELITE
CHAINS(Wo) SLIDES
I's
I
a'
5 6
SLIDES
4's 4's
I a's
4
as4'
a'
8's
7
*8 .9 TIME (GENERATIONS)
FIG. 2. Protocol of a membrane elution-Methocel segregation experiment. A total of 100 ml of a glucoseMOPS culture of E. coli B/rK was labeled for 4 min with 3H]thymidine (60.83 Ci/mmol; 10 ,Ci/ml) and filtered onto a Millipore membrane (type GS, 142-mm diameter). After the filter was washed with MOPSglucose medium, it was inverted and fresh medium was pumped through it (6). The first 4 min of elution yields the "wash-off," which is the ceUs not bound to the membrane. These are unselected cells and are grown for two to three generations before placement in Methocel so that cells with only one labeled strand are produced. Similarly, cells eluted after one or two generations of growth on the membrane are allowed to incubate in liquid for an additional period to assure the complete segregation of DNA strands. From the third generation of elution onward, there is no need for additional incubation in liquid culture as growth on the membrane is sufficient to allow the segregation of the DNA strands. After six generations of elution, the cells from the membrane are washed off with a rubber policeman and fresh medium, and the cells so obtained are placed on slides for chain formation. C + D and C + D + T are the times (6) at which decreases in radioactivity of less than and more than a factor of two, respectively, wiU occur according to both the PZ and SI models (1, 7).
mally present in liquid media) or the membrane binding technique (that may not release cells, as suggested in Fig. 1) may affect and weaken the critical experiment presented above. These possibilities appear unlikely for the following two reasons. First, our experiments are designed to detect differences and trends in segregation by using the Methocel technique throughout, thus minimizing the effect of any uniform error. It would be expected that, if the Methocel technique altered the segregation values in any particular sample of cells, one would still be able to observe any gross changes or trends predicted by the PZ model. Second, evidence that the membrane elution technique is working as expected is given by the variation in the elution
plateau ratios as described by Pierucci and Helmstetter (6) and as also observed by us in these experiments. First and second segregation. After completion of the synthesis of a new DNA strand, it will segregate into one of two daughter cells at the next cell division. What will happen if the pole toward which the new strand segregated is also preferentially chosen as the pole toward which the strand will segregate at the next cell division? Since the first segregation occurred with, presumably, an associated older strand that was the template for the new strand, at the second division this older strand must segregate away from the new strand and toward the pole to which it did not segregate at the first division.
DNA SEGREGATION IN E. COLI
VoL- 134, 1978
TABLE 3. Comparison of experimental and predicted segregation ratios for membrane elution experiment in glucose at 370C Calculated R value
Sample of cells
analyzed
Experimental PZ
PZmodel 3.0
SI model (P
R value
-0.75)a
3.0 2.84 Wash-off (unselected) 3.0 3.0 Generation 1 2.53 eluate 2.3 3.0 Generation 2 2.92 eluate 5.0 3.0 2.72 Generation 3 eluate 1.0 3.0 Generation 4 3.05 eluate 1.0 3.0 Generation 5 2.93 eluate 1.0 3.0 2.71 Generation 6 eluate 33.0 3.0 Membrane 3.07 a P is the probability that a strand will segregate toward the same pole that it segregated to in the previous generation.
This paradoxical situation means that all older strands will actually segregate preferentially toward the inner cells of chains rather than toward the outer cells of chains as observed. Thus, we must conclude that in any situation in which nonrandom segregation is taking place there must be some "seniority" system in which the older strands segregate preferentially toward the pole to which they segregated at the previous division. In the PZ model this is stated as the requirement that "attachment" to a pole does not occur until the newly synthesized strand is first used as a template. We merely wish to show that this type of rule (that attachment to a pole occurs when the strand is first used as a template) must be present in any system of nonrandom segregation of complementary strands, and therefore this type of strand movement is incor-
65
porated into the SI model (1). Mechanism of segregation. None of the results reported here bears directly on the mechanism of nonrandom segregation. However, if segregation is dependent on some DNA-membrane association, then the variable nonrandom segregation may be due to either the variation in the attachment of the DNA to the membrane or the variation of the association of the membrane to a particular pole of the bacterial cell
(1). ACKNOWLEDGMENT This work was supported by grant GB 40099 from the National Science Foundation.
LITERATURE CITED 1. Cooper, S., and M. Weinberger. 1977. Medium-dependent variation of deoxyribonucleic acid segregation in Escherichia coli. J. Bacteriol. 130:118-127. 2. Helmstetter, C. E., and 0. Pierucci. 1976. DNA synthesis during the division cycle of three substrains of Escherichia coli B/r. J. Mol. Biol. 102:477-486. 3. Lark, K. G., EL Eberle, R. A. Consgli, EL C. Minocha, N. Chai, and C. Lark. 1967. Chromosome segregation and the regulation of DNA replication, p. 63-89. In H. J. Vogel, J. 0. Lampen, and V. Bryson (ed.), Organizational biosynthesis. Academic Press Inc., New York. 4. Lin, E. C. C., Y. Hirota, and F. Jacob. 1971. On the process of cellular division in Escherichia coli. VI. Use of a Methocel-autoradiographic method for the study of cellular division in Escherichia coli. J. Bacteriol. 108:375-385. 5. Neidhardt, F. C., P. L. Bloch, and D. F. Smith. 1974. Culture medium for enterobacteria. J. Bacteriol. 119:736-747. 6. Pierucci, O., and C. E. Helmstetter. 1976. Chromosome segregation in Escherichia coli B/r at various growth rates. J. Bacteriol. 128:708-716. 7. Pierucci, O., and C. Zuchow8ki. 1973. Non-random segregation of DNA strands in Escherichia coli B/r. J. Mol. Biol. 80:477-503. 8. Ryter, A., Y. Hirota, and F. Jacob. 1968. DNA-membrane complex and nuolear segregation in bacteria. Cold Spring Harbor Symp. Quant. Biol. 33:669-676.
