Search tips
Search criteria 


Logo of nihpaAbout Author manuscriptsSubmit a manuscriptHHS Public Access; Author Manuscript; Accepted for publication in peer reviewed journal;
Mol Microbiol. Author manuscript; available in PMC 2010 November 4.
Published in final edited form as:
PMCID: PMC2973562

The Escherichia coli baby cell column: a novel cell synchronization method provides new insight into the bacterial cell cycle


We describe a new method for synchronizing bacterial cells. Cells that have transiently expressed an inducible mutant ‘sticky’ flagellin are adhered to a volume of glass beads suspended in a chromatography column though which growth medium is pumped. Following repression of flagellin synthesis, newborn cells are eluted from the column in large quantities exceeding that of current baby machine techniques by approximately 10-fold. Eluted cultures of ‘baby cells’ are highly synchronous as determined by analysis of DNA replication, cell division and other events, over time after elution from the column. We also show that use of ‘minutes after elution’ as a time metric permits much greater temporal resolution among sequential chromosomal events than the commonly used metric of cell size (length). The former approach reveals the existence of transient intermediate stages that are missed by the latter approach. This finding has two important implications. First, at a practical level, the baby cell column is a particularly powerful method for temporal analysis. Second, at a conceptual level, replication-related events are more tightly linked to cell birth (i.e. cell division) than to absolute cell mass.


Bacterial cells growing exponentially in batch culture are asynchronous, with different stages of the cell cycle represented in proportion to their relative durations. Investigations of the timing of events during the cell cycle have therefore often relied upon the generation of synchronous populations, which can then be followed over time (e.g. Cassler et al., 1995). Additionally, cytological studies have utilized asynchronous populations, with cell length (which increases linearly with time after cell birth) providing an approximate time line (e.g. Cook et al., 1987).

Synchronous populations have been obtained in either of two ways. In one approach, cells are arrested at a particular point, allowed to accumulate at that point, and then released from arrest. For example, nutrient starvation can be used to arrest cells in a relatively uniform state (Barner and Cohen, 1956). Also, DNA replication can be synchronized at the step of initiation by the use of temperature- sensitive alleles of dnaA or dnaC, whose products are specifically required at this point (Carl, 1970). Arrest methods can give extremely good synchrony, but have the undesirable effect of disturbing the natural physiology of the cells and thus are not ideal for analysis of basic cell cycle events under normal growth conditions.

A second approach, developed by Helmstetter and Cummings (1963), involves the specific collection of newly divided cells (newborn) by a physical method known as a ‘baby machine’. Cells from an exponentially growing culture are affixed to the bottom surface of a nitrocellulose membrane. As growth medium then flows through the membrane, attached cells grow and divide. As each dividing cell yields one newly formed daughter cell that is not membrane-attached, these newborn daughters are efficiently shed and can be collected. Helmstetter’s machine has gone through several adaptations, the most prominent being the addition of an adhesive (poly-D-lysine) to the membrane to allow the use of K-12 strains of Escherichia coli (Helmstetter et al., 1992; the B/r strains used initially attach naturally to the membrane). The newborn cell populations thus produced are unperturbed and very synchronous. Analyses of such baby cell populations have led, in great part, to our current understanding of the bacterial cell cycle (reviewed in Helmstetter, 1996).

The baby cell machine does, however, have one significant drawback: the yield of cells is quite low as compared with the amounts required for molecular experiments such as Southern blotting or microarray analysis. Such studies are therefore extremely difficult, requiring collection and pooling of multiple samples (e.g. Cassler et al., 1995). To address this limitation, we have developed a new method for collecting newborn cells, which we call a ‘baby cell column’. In this method, exponentially growing cells are attached to a glass bead column via their flagella. Growth medium is pumped through the column, with flagellin synthesis now turned off. Newly divided cells that lack flagella are released from the column and can be collected in the flow-through. A sample collected over a sufficiently small time interval contains a highly uniform population of newborn cells that can be examined directly and/or further incubated and monitored at another time(s) during the cell cycle. In this report we describe the operation of the baby cell column, document the yield and synchrony of the resulting newborn cell populations and compare results obtained with the baby cell column with those obtained using cell length as an index of time after birth. The latter comparison reveals that the baby cell column provides much higher temporal resolution. Implications of this finding as well as potential improvements and extensions of the method are discussed.

Results and discussion

Synchronization principle and design of the baby cell column

In our synchronization system, exponentially growing cells are attached to microscopic glass beads that are packed within a chromatography column (Figs 1A and and2;2; below). Growth medium is pumped through the beads, allowing attached cells to continue to grow and divide as in a normal batch culture. Because of specific features of the system, described below, each attached dividing cell yields one daughter that remains attached to its bead and one daughter that exits the column. An aliquot of released cells collected over a suitably short time interval provides a highly synchronous population of newborn cells, which can be further incubated and/or analysed.

Fig. 1
Cell attachment and synchronization principle
Fig. 2
Baby cell column

This system requires both strong attachment of loaded cells to the growth surface and selective lack of attachment of newborn cells. For attachment, cells are anchored to the glass beads via their flagella. E. coli flagella have an intrinsic tendency to stick to glass. This tendency was further enhanced by use of a special ‘sticky’ allele of the flagellin gene, fliCst (Fig. 1B), previously used to tether cells onto the surface of glass microscope slides for examination of flagellar movement (Scharf et al., 1998). When mixed with clean, uncoated glass beads, fliCst cells spontaneously attach themselves to the bead surface.

For specific release of newborn cells, flagellar synthesis was manipulated. Expression of the fliCst gene was placed under control of the isopropyl β-D-thiogalactopyranoside (IPTG)-inducible tac promoter, with a cognate lacIQ repressor present to give particularly tight repression (Fig. 1B; Experimental procedures). Cells are pregrown in the absence of IPTG and then, shortly before the start of the experiment, subjected to a brief pulse of IPTG to transiently induce fliCst expression. This pulse appears to result in the development of one or two flagella in most cells, as suggested by microscopic inspection (data not shown). Given this situation, an attached cell will release a newborn cell lacking flagella at the first or second division on the column. After the induction pulse, cells are attached to a volume of glass beads and then introduced into the column. IPTG is not present in the growth medium used to elute the cells. Use of a pulse of flagellin synthesis is essential to proper functioning of the column; if fliCst expression is constitutive, the initial yield of baby cells is very low, and after two to three generation times the column becomes clogged, as revealed by reduction in flow (data not shown). These effects are presumably attributed to reattachment of flagellated newborn cells to the glass beads. Thus, in our optimal protocol, unwanted overflagellation of cells is avoided by inducing flagellin synthesis in a short pulse, with prior and ensuing repression of the fliCst promoter with the hyperactive lacIq repressor.

The baby cell column (Fig. 2) utilizes a standard chromatography set-up with one minor specialized modification. In standard gel chromatography, resistance of the gel bed to the flow of buffer usually generates even (plug) flow of liquid throughout the horizontal plane of the column. However, glass beads comprise a highly porous bed, resulting in a more parabolic flow, i.e. the medium tends to run straight down the centre of the column. To correct for this situation, a 0.2 μm-pore membrane is fitted onto the top adapter, providing a resistance to flow that forces the medium to spread radially before moving down the column (Fig. 2B). In our standard the set-up, ~8 ml of cell-coated beads are loaded into a 16 mm chromatography column, which is then attached to a constant-flow peristaltic pump delivering medium at 5 ml min−1. The medium in the reservoir is aerated by brisk stirring. The entire assembly is placed into a constant temperature room to maintain growth temperature. The components for the system are readily available, e.g. off-the-shelf small diameter glass beads, a standard chromatography column and routine laboratory accessories (Fig. 2; Experimental procedures).

Characterization of column yield

To document the basic behaviour of the system, a baby cell column was run in AB proline medium at 30°C, which supports a doubling time of ~153 min for the CM735-derived strain used in this analysis. Under these conditions, cells grow with a linear, ‘eukaryotic-like’ cell cycle, with sequential G1, S, G2 and ‘M’ phases (i.e. with no more than a single round of replication occurring at any given time). Such conditions allow for straightforward analysis of cell cycle events (below; Bates and Kleckner, 2005). Beads were loaded into the column and elution was started at 5 ml min−1. Immediately, a 1 min sample was taken and then again at 5 min intervals thereafter. As 1 min represents 1/153 = 0.65% of the length of a single cell cycle, the sampled cells should all be at very similar stages in the cell cycle.

Column yield over time

The number of cells per unit eluted volume was determined in each of the successively collected samples. Cell yield was initially ~8 × 109 cells min−1 but by t = 40 min dropped to ~2.5 × 108 cells min−1 and then remained relatively steady for two or more division times, excluding minor fluctuations (Fig. 3A; below). The initial adjustment period prior to t = 40 min was always seen and likely represents elution of cells that were not tightly attached to beads.

Fig. 3
Column yield over time

Minor fluctuations

Once the column achieved steady state, two interesting minor fluctuations were detected. First, yields were seen to vary by 30%, in cyclic fashion, with a periodicity of ~150 min, which corresponds closely to the doubling time of cells grown in batch culture under similar conditions (153 min; Fig. 3A). This cyclic variation is explained by the fact that an exponentially growing batch culture, such as the one used here to attach to the glass beads, exhibits an age distribution of two newborn cells (just divided) for every dividing cell (just about to divide) (Fig. 3A, inset). Upon addition of the starting exponential culture to the glass beads (about 6 min prior to taking the first sample), all cells became attached to the beads essentially at the same time, and the cells in this attached culture proceeded through the cell cycle at a constant rate. Given the newborn-skewed age distribution of the starting culture, the number of attached cells in the process of division (and thus the number of eluted cells) increased as the attached culture matured, to a maximum exactly one generation time (~153 min) after the initial attachment (~t = −6 min). Thus, the number of eluted cells experienced peaks at generation intervals after attachment (Fig 3A, t = ~150 and t = ~310 min). If the culture attached to the column comprised a perfect exponential population (Fig. 3A, inset), the baby cell yield would exhibit a saw-tooth pattern as shown (Fig. 3A, theoretical curve; Experimental procedures). The closeness to which the experimental curve matches the theoretical curve implies that the attached cells are growing exponentially, without perturbation, through at least two full division cycles.

A second feature, underlying these fluctuations, is a slow, steady increase in total cell yield as shown by a corresponding increase in the maximum cell yield during the second division cycle as compared with the first (5.7 × 108 and 4.6 × 108 cells min−1 respectively). This characteristic suggests that the column-attached population is growing in size as the column ages, presumably because some newborn cells retain flagella and remain attached to the column rather than eluting. The observed increase in column yield, of about 20% over two generations, would suggest that about 10% of all newborn cells reattached to the column. The cell yield did not further increase after about three generations on the column (data not shown). This feature suggests that the increase observed during earlier generations reflects the reattachment of cells that carried residual flagella produced during the induction period, with no further increase after these flagella were diluted out.

To further examine these fluctuations, we performed a control experiment in which beads were added to the exponential culture in 10 equal (1 g) increments every 15 min, such that cell attachment was spaced out over a period of 150 min. In this protocol, small peaks of newborn cells would be expected every 10 min, which, at the resolution of these experiments, would result in an essentially constant yield of baby cells over time. As expected, after an initial adjustment period, cell yields were generally steady at ~3–4 × 108 cells min−1 and no regular periodic fluctuations could be detected; moreover, a slow gradual increase in total cell yield was again observed, in accord with the envisioned general effect (Fig. 3B, red line).


Once the column reaches steady state, the number of cells released by the column during a single generation time corresponds to the number that are both stably affixed to the column and are actively dividing to release newborn cells. For the two generations shown by the data in Fig. 3A, this value is 5.6 × 1010 and 6.8 × 1010 respectively. These values are very similar to the total number of cells that are affixed to the column (i.e. the number of cells initially loaded in the column minus the number of all cells eluted through the first 20 min ‘adjustment period’; Experimental procedures). We infer that essentially all of the cells remaining on the column after the adjustment period are actively dividing and releasing newborn cells.


The yield of newborn cells per minute collection time achieved by this method is 3–5 × 108 per 1 min sample (Fig. 3A). This is about 10-fold higher than would be achieved by the Helmstetter baby machine under comparable conditions (for comparison, see Helmstetter et al., 1992). Improvement results from the use of glass beads which, being both three-dimensional and relatively small (180 μm in diameter), present a large total surface area. In our standard set-up, the 10 g volume of beads creates ~1.7 × 103 cm2 of growth surface. Moreover, these beads are contained in a total volume of about 8 ml, thus allowing a very high number of cells to be grown in a small volume of medium.

Effective column lifetime

We normally begin collecting cell samples after two generation times, i.e. at t = ~300 min under these conditions, to ensure that all loosely attached cells have been removed, and elution rates have stabilized. The column continued to exhibit the properties described above and cells continued to exhibit good synchrony with respect to all assayed events for at least two doubling times thereafter, i.e. until ~6 h after assembly and for a total of approximately four generation times (below; data not shown). After this point, column yields tended to decay, slowly until t = 9 h and then significantly after 10 h, with collected cells also exhibiting significant loss of synchrony at these late times (data not shown). Deterioration of the column might be attributable either to eventual detachment of flagellated cells (via loss of contact or flagellum breakage) or, deterioration of growth conditions, leading to a less-productive population and/or cell death. We favour the former possibility because cells collected from a 10 h old column exhibited their first doubling at the same time as cells collected from the column soon after it was assembled (data not shown). Had growth conditions deteriorated on the column, an adjustment period prior to the next division would have been expected, with a corresponding delay and/or increased asynchrony in the timing of the next division.

Degree of synchrony of DNA replication in baby cell populations

As one way of assessing the synchrony of cell populations produced by the baby cell column, the timing of DNA replication was examined. A column was assembled under the conditions described above and run undisturbed for 5.5 h. A series of 1 min samples were collected in rapid succession and grown for different amounts of time in the elution medium without further dilution (~3 × 108 cells in 5 ml) at 30°C, thus giving a series of sample populations representing different times after elution from the column. This approach provided a larger number of cells for analysis at each time point than would have been achieved by taking only a single elution sample and removing aliquots at different appropriate times after collection. While not necessary for all analyses, this protocol is useful for some types of experiments, because it permits an unlimited number of time points, each of which can be analysed by a number of different types of assays.

For the current purpose, each sample was analysed for DNA content by flow cytometry, and the fraction of cells exhibiting DNA contents intermediate between 1 (G1 peak) and 2 (G2M peak) chromosomes was determined from the resulting distributions (Fig. 4A, S peak; Experimental procedures). In an asynchronous batch culture grown under similar conditions, 24% of cells exhibited 1-chromosome equivalent of DNA, 39% exhibited 2-chromosome equivalents of DNA, and the remaining 37% exhibited intermediate DNA contents, implying that they were in the process of bulk DNA replication (Fig. 4B, left panel). As expected, almost no cells exhibited more than two chromosomes at this growth rate. The sample of cells that had just emerged from the baby cell column exhibits a very different distribution (Fig. 4B, middle panel): 84% of cells contained a single chromosome equivalent, implying that the large majority of cells had not yet initiated replication and thus were in G1. Of the remaining cells, 9% exhibited two chromosome equivalents and 7% exhibited intermediate DNA contents. Once again, essentially no cells exhibited more than two chromosome equivalents of DNA. Progression of DNA replication was then monitored by determining the percentage of cells showing intermediate DNA contents in the entire series of eluted cultures (Fig. 4C). Cells undergoing DNA replication appeared in two dramatic waves that were appropriately separated by ~150 min, the length of a single cell cycle.

Fig. 4
Synchrony of baby cell cultures

The area under any such primary curve is proportional to the amount of time that cells spend in the corresponding stage; for a long-lived stage, cells accumulate for a long time and then disappear, whereas for a transient stage, very few cells are ever present in that stage at any given time and the corresponding peak is very small. The lifespan of a stage is given quantitatively by determining the area under the curve (which has units of percentage cells × time) and dividing by the percentage of cells that progress through this stage, to give a duration for the stage (in units of time). If all cells in an eluted sample were behaving as a single synchronous population, the perentage of cells progressing through each stage would be 100%. However, there appears to be a subset of 15–20% of cells that are not part of the main synchronous population, which comprises each primary curve. For example, in the current experiment, ~16% of cells in the t = 0 sample contained more than one chromosome equivalent of DNA. A similar level of ‘inappropriate’ cells is seen for any stage examined (Bates and Kleckner, 2005). We infer that these cells represent inefficiencies in cell release and/or elution and that they can be considered as a separate ‘subpopulation’ of essentially asynchronously growing cells. This asynchronous population must be excluded from the calculation of lifespans. Thus, in the current experiment, the areas under the primary curves (Fig. 4C) were divided by 84%.

The calculated lifespans of the DNA replication periods in the two rounds of replication analysed in Fig. 4 were 47 min and 49 min respectively. There is little published data available for cell cycle parameters in E. coli K-12 strains at 30°C. However, the values obtained here are qualitatively reasonable, as they are within 30% of earlier measurements obtained by flow cytometry analysis of cells of the same background growing at a comparable doubling time but at 37°C (64 min; Allman et al., 1991).

Additional information can be obtained by integrating under the primary non-cumulative curves (Fig. 4D; Experimental procedures). The resulting cumulative curves describe the percentage of cells that had entered DNA replication as a function of time for each wave. Parallel curves displaced by the lifespan of the corresponding period describe the percentage of cells that had exited DNA replication as a function of time. Cumulative curves for the two successive rounds of DNA replication revealed that they were separated by almost exactly 150 min, as given by the interval between the times at which 50% of cells had entered replication in the two waves (Fig. 4D).

Cumulative curves also describe the degree of synchrony of the majority cell population: the steeper the slope the more synchronous the culture. In a culture that was perfectly synchronized (100% of eluted cells are newborn), and in which replication initiation occurred at the same moment after birth in every cell, the cumulative curve would be vertical. In the experiment described here, cells entered the first replication period at the rate of about 10% of cells per 5 min, with 50% of cells entering within a 30 min span (Fig. 4D). Strikingly, the cumulative curve for the second wave of replication was only slightly less steep than that for the first wave, implying that the same high degree of synchrony was maintained through at least two cycles. Even tighter synchrony, e.g. with 50% of cells entering a stage in as few as 15 min, has been observed in some experiments (e.g. Bates and Kleckner, 2005).

Analysis of cell-length distributions in synchronous cultures

The cell samples analysed for DNA replication in the previous section were also analysed for cell lengths, determined by microscopic observation of 100 cells at each time point (Experimental procedures). In the newborn culture (Fig. 4E, bottom panel) mean cell length was significantly smaller, and cell lengths were more tightly distributed (mean = 2.1 μm, SD = 0.3 μm) than in an asynchronous culture grown under similar conditions (top panel; mean = 2.9 μm, SD = 1.1 μm).

Overall, cell mass, and thus cell length, increase linearly with time in the cell cycle, doubling from the time of birth to the time of the next cell division, at which point both mass and length decrease by half. This progression is clearly visible in the baby cell populations (Fig. 4F): average cell length increased from a mean of 2.1 μm at t = 0 to a maximum of 4.5 μm at 150 min. At 180 min, the average cell length abruptly dropped to 2.3 μm, indicating that most cells in the culture divided between 150 and 180 min, in accord with the 153 min doubling time of this strain in batch culture. These results further illustrate the overall synchrony of cells in samples from the baby cell column.

Time after cell birth versus cell mass as a time metric

As cell length increases essentially linearly with time during the cell cycle, many previous studies have analysed the relative and absolute timing of events by using cell length as a time metric. In the baby cell column approach, the time metric is ‘time after emergence from the column’, which corresponds by definition to the time after the previous cell division or ‘birth’. We were interested to compare temporal resolution provided by the two approaches. For this purpose, we examined the appearance and disappearance of cytologically detectable foci representing the replication origin, oriC, as revealed by fluorescence in situ hybridization (FISH) and of foci representing complexes of DnaX protein, the beta-clamp subunit of the replicative polymerase of E. coli, as revealed by immunostaining of a DnaX-GFP fusion protein. Primary data were provided by a study described in detail elsewhere (Bates and Kleckner, 2005).

When the cell types were plotted as a function of time after elution from the column, a clear progression was revealed (Fig. 5A): cells were born with a single origin focus and no DnaX foci (1,0; open circles); a single DnaX focus then appeared (1,1; light grey circles); origins then separated (2,1-A; red circles, left peak); left and right replisomes and their corresponding DnaX foci then separated (2,2; dark grey circles); one DnaX focus then disappeared (2,1-B; red circles, right peak) and the second DnaX focus disappeared (2,0; green circles). Early classes then reappeared as the cells entered the next cell cycle. The (1,1) (2,2) and (2,0) classes were prominent, implying that they were long-lived; the (1,0) and the two (2,1) classes were less prominent, implying that they were more transient. These same differences are also reflected quantitatively in the lifespans of the different stages (figure legend of Fig. 5). For cells in the final stage (2,0), an appropriate cumulative curve showed that 50% of cells entered this stage within a span of ~21 min (Fig. 5A, dashed green line), implying a satisfactory high degree of synchrony in this particular experiment. Additional data suggest that the baby cell column permits temporal resolution of events that are separated by an interval as short as 5 min (Bates and Kleckner, 2005).

Fig. 5
Evaluation of cell length as a determinant of cell age

For comparison, all of the cells in all time points were pooled and binned according to cell length (Fig. 5B, top). Then, for each bin, the percentage of cells having each of the six (origin, DnaX) phenotypes was plotted as a function of the mean cell length of that bin (Fig. 5B, bottom). The general tendency for the prominent (1,1) (2,2) and (2,0) classes to occur in succession can again be seen. However, stages and relationships involving the three more transient stages are not discernible: the relationship between the transient (1,0) stage and the ensuing (1,1) stage is obscured and (2,1) cells are so broadly distributed at such a low level that no specific peak can be defined. These findings show that use of the baby cell column permits higher temporal resolution of sequential events, making it possible to reveal stages that are too closely spaced and/or too transient to be detected in asynchronous cultures with cell length as the time metric.

Timing of replication and related events is more tightly coupled to the prior cell division than to cell mass

These findings also have a further, more fundamental, implication: as cell length is exactly indicative of cell mass, our results imply that the timing of origin separation and DnaX appearance and disappearance are more tightly linked to the timing of the prior cell division than to absolute cell mass. Moreover, additional studies, involving comparisons of timing at different growth rates, further suggest that the assayed origin and DnaX events are linked to the DNA replication cycle (Bates and Kleckner, 2005). This finding is interesting because it has often been proposed that replication initiation is linked specifically to absolute cell mass (e.g. Donachie and Blakely, 2003). Our results would suggest, in contrast, that replication initiation is more tightly temporally coupled to the prior cell division than to cell mass.

To provide more direct evidence for this possibility, we analysed the available data in another way, using cumulative curves. The steepness of a particular cumulative curve reflects not only degree of synchrony within the population but also uniformity in the time of occurrence of an event following division (above). Thus, for a given population, two events whose timing is more or less tightly linked to the prior cell division will show, respectively, steeper or shallower cumulative curves. The cumulative curves for onset of DNA replication are quite steep, with 50% of cells entering S phase within a 29 min span in both the first and second rounds (Fig. 5C, bottom, black curves). In fact, the cumulative curves for most events analysed thus far exhibit very similar slopes (Bates and Kleckner, 2005) as seen not only for DNA replication but also for the origin and replisome changes described above, other chromosomal events, and division-related events such as appearance of a visible septum and the occurrence of division itself (Bates and Kleckner, 2005). By implication, all of these events occur with similarly tight coupling to the preceding cell division. To similarly assess the linkage of cell mass to the timing of the prior division, we derived a cumulative curve for entry of cells into the ‘stage’ at which they had reached a length of 3 μm or longer (3 μm was chosen arbitrarily; it represents the average length of a cell that is about halfway through the cell cycle). The primary curve shows that cells accumulate in this ≥3 μm stage more gradually than they accumulate in the DNA replication stage and then exit at the time of cell division (~153 min; Fig. 5C, top; gold curve). Correspondingly, the cumulative curve for this stage has a much shallower slope than that for initiation of DNA replication, with 50% of cells achieving 3 μm in length within a 58 min span (Fig. 5C, bottom; gold curve). The cumulative curve for the second round of DNA replication (after cell division) is nearly as steep as the first round, therefore the shallow slope of the ≥3 μm curve is not attributed to a deterioration of synchrony in the cell populations (Fig. 5C, bottom). We conclude that the time of initiation of DNA replication is more tightly coupled to the time of the prior division than is the time at which the cell reaches a particular length (mass). This possibility is also supported by other lines of evidence and is in accord with the predictions of a new model for control of replication initiation in which the act of cell division licenses the occurrence of the next round of replication initiation (Bates and Kleckner, 2005).

Extensions and future prospects

Scaling up

One advantage of using glass beads as a matrix for cell attachment is the ease with which the scale of the system can be increased. In theory, scale is limited only by the capacity of the column and the ability to pump medium through the column at a sufficiently high rate, with appropriate redistribution across the upper surface of the bead matrix. Scaling up will be advantageous for some types of analyses. The set-up described here produces ~3 × 108 cells min−1 in a 5 ml volume (above). However, some experiments, e.g. DNA footprinting, require very large amounts of cells (~109, Cassler et al., 1995), thus necessitating pooling of several samples even at the relatively high current yield. Preliminary experiments have been carried out using a column of 5 cm in diameter and 50 g of glass beads, as compared to the 1.6 cm column and 10 g of beads used in the current set-up. The larger assembly resulted in a two- to threefold increase in cell yield per minute of collection time that, while still an improvement, was somewhat less than the factor of five that might have been expected. Resulting cells were still highly synchronous, but were slightly less synchronous than seen with the smaller column as judged by flow cytometry analysis. Both of these deficiencies are attributable to inefficient flow of medium through the wider column and we are currently working on a solution to this problem.

Analysis of fast growth rates

The baby cell column can be run with virtually any type of growth medium. This includes rich media such as Luria–Bertani broth, although flow rates must be increased substantially under these conditions to maintain steady-state growth of attached cells (data not shown). This requirement is likely attributed to a greater need to exchange spent media within the column, which contains less oxygen and a higher concentration of metabolic wastes under conditions of fast growth. The flexibility of the baby cell column with respect to growth medium also represents an additional improvement over the Helmstetter baby machine, because the poly-D-lysine-mediated attachment of cells to nitrocellulose membranes used in that method is highly sensitive to the presence of charged molecules in the media, such as casamino acids (Helmstetter and Cummings, 1964) and thus is not so suitable for analysis of fast growth rates, at least in E. coli K-12 where this attachment method is required (Introduction). It will be very interesting to understand the relationship between patterns observed under slow-growth (linear cell cycle) conditions (e.g. Bates and Kleckner, 2005) and events under rapid-growth conditions, which result in multiple overlapping replication cycles.

Applicability to other strains of E. coli and to other bacteria

The baby cell column should be easily applicable to any strain of E. coli. The only requirements are: (i) presence of a complete and fully functional flagellar assembly, which can be determined by a standard motility assay (Berg, 2004); (ii) integration of the Ptac-fliCst; lacIQ construct somewhere in the genome, which can be accomplished by homologous recombination without the need for linked antibiotic resistance markers (Experimental procedures); and (iii) absence of any untoward tendency for ‘stickiness’ of the overall cell surface. We have performed successful baby cell experiments in both CM735 (this study; Bates and Kleckner, 2005) and W1485 strain backgrounds, which were chosen primarily because they exhibit particularly good motility. Additionally, both strains perform well in flow cytometry experiments for analysis of DNA replication (e.g. Allman et al., 1991). Most other common laboratory strains (e.g. MG1655) also meet the above requirement and thus should be amenable to analysis via the baby cell column. This approach should be equally applicable to relatives of E. coli (e.g. Salmonella typhimurium) that are flagellated and exhibit common requirements for gene expression. Where necessary, external stickiness attributed to extracellular peptidoglycan can be eliminated by introduction of a gal-mutation; in fact, homologous integration of the Ptac-fliCst; lacIQ construct has exactly this effect.

Possible extension to other organisms

It is intriguing to consider the possibility that an analogous approach, involving elution from a glass bead column, might also be applicable to microorganisms without flagella or to other types of cells in dispersed cell culture by using an alternative method for attachment of the starting population. For example, cells marked with one or two antibodies to a cell surface component could be attached to beads coated with an appropriate secondary antibody. Dilution of the primary antibody by cell growth would yield newborn cells that could no longer attach to the column. Such ‘crossover’ applications have been demonstrated for the Helmstetter baby machine, which has been applied to yeast (Helmstetter, 1991) and, more recently, even to mammalian cells (Thornton et al., 2002).

Experimental procedures

Bacterial strains

The strain used for baby cell experiments, NK9386, is CM735 (Hansen and von Meyenburg, 1979) carrying the sticky flagellin (fliCst) tethering construct inserted at the galE locus (Fig. 1B). The endogenous fliC sequence was disrupted without affecting expression of other genes in the fli operon (critical for flagellar biosynthesis). Both modifications were made on gene replacement vectors, and then transferred onto the chromosome by homologous recombination in a selectable two-step integration-excision process. Construction details follow.

To disrupt the fliC gene, the fliD-fliC-fliA sequence was amplified by polymerase chain reaction (PCR) in two steps, leaving a 1.5 kb in frame deletion within the fliC gene. The two products were then ligated together and inserted into the gene replacement vector pBIP (Slater and Maurer, 1993) between the ApaI and SacI sites, resulting in pNK3874. This plasmid was then used to transfer the fliC deletion onto the chromosome of CM735, resulting in NK9375.

To make the Ptac:: fliCst tethering construct, the 1.7 kb HinC II fragment from pFD313 (Kuwajima, 1988; Scharf et al., 1998) containing the original fliCst in frame deletion mutation (57 bp deletion with 18 bp linker insertion) was cloned into an IPTG-inducible Ptac lacIQ expression plasmid, pNK3868. The hyperactive lacIQ repressor gene (Wang et al., 1983), which strongly suppresses Ptac in the un-induced state, is critical for proper baby cell column operation (Results). galE flanking sequences of 200 bp (for integration purposes) were generated by PCR and added to both ends of the Ptac::fliCst lacIQ sequence. Finally, the entire galE-Ptac::fliCst-lacIQ -galE′ sequence was inserted into the gene transfer vector, pKO3 (Link et al., 1997), at the SmaI site, resulting in pNK3886. This plasmid was then used to insert the tethering sequence into NK9375 at the galE locus by homologous recombination, resulting in NK9386.

Baby cell column

An XK-16 chromatography column (Pharmacia) fitted with two flow adapters served as the growth chamber. Growth medium entered the column through the top flow adapter, to which was attached approximately 3 feet of capillary tubing (1.2 mm I.D. ETFE tubing, Pharmacia). The tubing was fed through a constant flow peristaltic pump (e.g. Rabbit pump manufactured by Rainin) and the free end was submerged in a sterile medium. Poor quality pumps that result in ‘pulsing’ should not be used. A 0.22 μm membrane (GV Durpapore, Millipore) prewetted with water was placed between the support screen and the net ring as shown (Fig. 2B). The membrane acts as a flow barrier, causing the medium to flow radially and rise in pressure, so that fluid is injected onto the column bed more evenly. Therefore, the membrane must sit between the net ring and the support screen with no holes or cracks. The bottom flow adapter was attached to a 25 cm length of capillary tubing and was fitted with a support screen and net ring (no 0.22 μm membrane). The flow adapters were screwed onto the column and the entire assembly was cleaned by running 70% ethanol and then distilled water through the column for 10 min each at a flow rate of 10 ml min−1. The prepared column was then placed into a 30°C constant-temperature room for at least 1 h before glass beads were added.

Cell synchronization

One litre of NK9386 cells were grown to mid-exponential phase (OD450 = 0.2), 0.1 mM IPTG was added, and the culture was incubated for an additional 1 h with moderate shaking (100 r.p.m. in a 2 l plastic flask; fliCst expressing cells have a tendency to stick to non-plastic surfaces and to each other, therefore physical manipulation was minimized to avoid shearing of flagella). The culture was concentrated 20-fold by centrifugation (3000 r.p.m. 5 min−1) and added to 10 g of glass beads (180 μm diameter, Sigma) that had been cleaned with ethanol and water (above), and warmed to 30°C. The mixture was then allowed to sit for 15 min without shaking at 30°C. The top flow adapter was removed from the baby column, and the cell/bead mixture was poured into the column. Working quickly before all the media drained from the column, the adapter was screwed on, lowered snugly to the surface of the beads, and sealed in place. Fresh (below), warm, aerated (via brisk stirring) medium without IPTG was delivered at a flow rate of 25 ml min−1 for 5 min to remove loosely attached cells. The flow rate was then reduced to 5 ml min−1 and allowed to run undisturbed for two doublings, or a minimum of 2 h under faster growth conditions. Samples were collected and checked for synchrony by microscopic observation (small homogeneous cells) and flow cytometry (clean 1 chromosome peak). Approximately 2 × 1011 cells were initially incubated with glass beads. Approximately 30% of these cells (~6 × 1010) actually attached to the beads and were incorporated into the column.

Total synchrony of eluted cultures is estimated to be ~79%. This figure is roughly equal to the percentage of cells containing a single unreplicated chromosome immediately after coming off the column (84%, Fig. 4B and C). However, this is a slight overestimation, because this number includes cells in the asynchronous population that also happen to be in the 1-chromosome stage. Assuming that the asynchronous population exhibits a log age distribution (Fig. 3A, inset), and given the timing of DNA replication (above), we estimate that these cells represent ~5% of the total population, resulting in ~79% total synchrony. Note, as this estimation is determined by synchrony in replication, any inherent variance in the timing of replication initiation among cells of the same age will artificially lower our estimate of overall culture synchrony.

In this study, fresh AB minimal medium (Clark and Maaloe, 1967) supplemented with thiamine and proline (0.2% ea), and the required amino acids, tryptophan, histidine and methionine (20 μg ml−1 ea) was used for both the batch-grown starting cultures and for use in the baby cell column. Trial experiments in which conditioned media (recovered and filtered from early exponential phase cultures) was used in place of fresh media for pumping through the column, showed no detectable differences in cell growth or cell cycle timing in eluted cultures (data not shown).

Microscopy and flow cytometry

To determine cell concentration, baby cell cultures were serially diluted, 5 μl aliquots were placed onto eight-well microscope slides that were precoated with poly-D-lysine (60K mol wt, Sigma Aldrich), and the wells were photographed by phase contrast microscopy using a Zeiss Axioplan2 microscope equipped with a Princeton CCD camera. The number of cells per well was then counted using a detection algorithm in the Metamorph software package, and this number (all eight wells were counted and averaged) was converted to cells per millilitre. To determine cell length, cells were photographed and measured using spheres of a known size (Molecular Probes) to calibrate distances. Approximately 100 cells were analysed and averaged per time point.

For quantification of cellular DNA content, 5 ml samples were spun down, resuspended in 100 μl ice-cold TE (10 mM Tris-Cl, pH 7.5, 1 mM EDTA), and fixed in 1 ml ice-cold 77% ethanol. For DNA staining, cells were washed twice in TE and resuspended in 250 μl TE with 200 μg ml−1 RNaseA, then incubated at room temperature for 15 min. Sytox Green (Molecular Probes) was added to 2.5 μM and cells were incubated for an additional 15 min. DNA fluorescence (500–550 nm) was measured using a FACScan flow cytometer (Becton Dickinson) for 2 × 104 individual cells per sample.

Because of the lack of separation between the 1-chromosome and 2-chromosome peaks, commercial flow cytometry analysis programs were unable to reproducibly identify the different cell fractions, and therefore a hand-method was devised (Bates and Kleckner, 2005). To calculate the percentage of cells in the process of DNA replication, the number of cells containing full genome compliments (either one or two fully replicated chromosomes) was subtracted from the total number of cells. As the DNA fluorescence histogram peaks corresponding to 1 and 2 chromosomes are partially obscured by cells in early and late stages of replication, these fractions were approximated by quantifying the ‘outside’ half of the corresponding peaks (the sides of the peaks that do not overlap cells in the process of replication), and multiplying by 2 (Fig. 4A). Five independent calculations were made for each histogram and the results were averaged.

Cumulative curve analysis

To more accurately estimate the time of start and end of a particular stage, the raw data curve is converted to a set of cumulative curves as follows (also see Hunter and Kleckner, 2001). The ends of the raw data curve (e.g. t = 0, 90 min for the first replication cycle) are first set to zero, so as not to include data outside of the stage of interest. This operation is done primarily to facilitate area computations, however, it also helps to exclude ‘coincident’ cells from a re-occurring stage (e.g. 2ori, 1DnaX stages in Fig. 5A). Next, the area under the raw data curve between each set of consecutive time points (% cells × time) is calculated using the half-height method, whereby the area of a rectangle with height half-way between the two time points is calculated. The total areas for the entire stage are then summed to give the total area. The stage lifespan is then calculated as the total area divided by the total percentage cells going through the stage (84%; Text). The raw data curve is then smoothed by an interpolation function (excel), and a cumulative curve is generated by summing consecutive data points from the smoothed (entry) curve. The exit curve is then generated by displacing the entry curve to the right by exactly one lifespan. Entry and exit times of the stage are given by the times at which the entry and exit curves reach 50%.

Theoretical baby column yield from an ideal culture

The yield from the baby cell column within a particular time interval is equal to the number of bound cells that divided during that interval. This number changes over time as the bound culture, which has a non-linear age distribution (below), progresses through cycles of cell division. For an ideal culture, growing exponentially with no cell-to-cell variation, the number of cells produced in any time interval is the product of the total number of cells bound to the column, the frequency of bound cells that actually divided during that interval, and the division rate of the culture in doublings per minute (Fig. 3A, theoretical curve).

The number of cells bound to the column was estimated to be ~6 × 1010, about 30% of the cells that were initially incubated with the beads. The frequency of bound cells from the ideal culture that divided during each time point interval is determined from the age distribution of a theoretical ideal population given by the following function:


where a is the age and n(a) is the probability density for a cell to be of age a (Powell, 1956; Lindmo, 1982). The division rate of the culture is 153 min, or 6.5 × 10−3 doublings per minute.


This work was supported by National Institutes of Health (NIH) Grant GM25326 to N.K. and NIH postdoctoral Grant GM20627 to D.B.


  • Allman R, Schjerven T, Boye E. Cell cycle parameters of Escherichia coli K-12. J Bacteriol. 1991;173:7970–7974. [PMC free article] [PubMed]
  • Barner HD, Cohen SS. Synchronization of division of a thymineless mutant of Escherichia coli. J Bacteriol. 1956;72:115–123. [PMC free article] [PubMed]
  • Bates D, Kleckner N. Chromosome and replisome dynamics in E. coli: loss of sister cohesion triggers global chromosome movement and mediates chromosome segregation. Cell. 2005 (in press) [PMC free article] [PubMed]
  • Berg HC. E. Coli in Motion. New York: Springer; 2004.
  • Carl PL. Escherichia coli mutants with temperature-sensitive synthesis of DNA. Mol Gen Genet. 1970;109:107–122. [PubMed]
  • Cassler MR, Grimwade JE, Leonard AC. Cell cycle-specific changes in nucleoprotein complexes at a chromosomal replication origin. EMBO J. 1995;14:5833–5841. [PubMed]
  • Clark DJ, Maaloe O. DNA replication and the division cycle of Escherichia coli B/r. J Mol Biol. 1967;23:99–112.
  • Cook WR, Kepes F, Joseleau-Petit D, MacAlister TJ, Rothfield LI. Proposed mechanism for generation and localization of new cell division sites during the division cycle of Escherichia coli. Proc Natl Acad Sci USA. 1987;84:7144–7148. [PubMed]
  • Donachie WD, Blakely GW. Coupling the initiation of chromosome replication to cell size in Escherichia coli. Curr Opin Microbiol. 2003;6:146–150. [PubMed]
  • Hansen FG, von Meyenburg K. Characterization of the dnaA, gyrB and other genes in the dnaA region of the Escherichia coli chromosome on specialized transducing phages lambda tna. Mol Gen Genet. 1979;175:135–144. [PubMed]
  • Helmstetter CE. Description of a baby machine for Saccharomyces cerevisiae. New Biol. 1991;3:1089–1096. [PubMed]
  • Helmstetter CE. Timing of synthetic activities in the cell cycle. In: Neidhardt FC, Curtiss R, Ingraham JL, Lin EC, Low KB, Magasanik B, et al., editors. Escherichia coli and Salmonella: Cellular and Molecular Biology. Vol. 2. Washington DC: American Society for Microbiology Press; 1996. pp. 1627–1639.
  • Helmstetter CE, Cummings DJ. Bacterial synchronization by selection of cells at division. Proc Natl Acad Sci USA. 1963;50:767–774. [PubMed]
  • Helmstetter CE, Cummings DJ. An improved method for the selection of bacterial cells at division. Biochim Biophys Acta. 1964;82:608–610. [PubMed]
  • Helmstetter CE, Eenhuis C, Theisen P, Grimwade J, Leonard AC. Improved bacterial baby machine: application to Escherichia coli K-12. J Bacteriol. 1992;174:3445–3449. [PMC free article] [PubMed]
  • Hunter N, Kleckner N. The single-end invasion: an asymmetric intermediate at the double-strand break to double-holliday junction transition of meiotic recombination. Cell. 2001;106:59–70. [PubMed]
  • Kuwajima G. Construction of a minimum-size functional flagellin of Escherichia coli. J Bacteriol. 1988;170:3305–3309. [PMC free article] [PubMed]
  • Lindmo T. Kinetics of protein and DNA synthesis studied by mathematical modelling of flow cytometric protein and DNA histograms. Cell Tissue Kinet. 1982;15:197–211. [PubMed]
  • Link AJ, Phillips D, Church GM. Methods for generating precise deletions and insertions in the genome of wild-type Escherichia coli: application to open reading frame characterization. J Bacteriol. 1997;179:6228–6237. [PMC free article] [PubMed]
  • Powell EO. Growth rate and generation time of bacteria, with special reference to continuous culture. J Gen Microbiol. 1956;15:492–511. [PubMed]
  • Scharf BE, Fahrner KA, Turner L, Berg HC. Control of direction of flagellar rotation in bacterial chemotaxis. Proc Natl Acad Sci USA. 1998;95:201–206. [PubMed]
  • Slater S, Maurer R. Simple phagemid-based system for generating allele replacements in Escherichia coli. J Bacteriol. 1993;175:4260–4262. [PMC free article] [PubMed]
  • Thornton M, Eward KL, Helmstetter CE. Production of minimally disturbed synchronous cultures of hematopoietic cells. Biotechniques. 2002;32:1098–1100. 1102–1105. [PubMed]
  • Wang JC, Peck LJ, Becherer K. DNA super-coiling and its effects on DNA structure and function. Cold Spring Harb Symp Quant Biol. 1983;47(Pt. 1):85–91. [PubMed]