Generation of cheZ knock-out strain by using λ-Red mediated deletion strategy

We first constructed a plasmid (PBSKcl) that contained a chloramphenicol (Cm) resistance module as a selectable marker. These were used as templates for the PCR amplification of DNA ‘targeting cassettes’, to delete the CDS of interest. PCR primers each included 45–50 bases homologous to the chromosome immediately upstream (forward primer, sense strand) and downstream (reverse primer, antisense strand) of cheZ gene, as well as 17 bases homologous to the plasmid template.
We introduced the λ-Red recombination plasmid (pSim6)[1] into MG1655, namely MA6 for our chromosomal deletion. In MA6, exogenous DNA was efficiently introduced by electroporation, and integrated into the chromosome via a homologous recombination-based process mediated by the Beta and Exo proteins, which are located in a defective cI857 λ-prophage. A schematic illustrating the procedure is shown in Fig.1A. After the cheZ gene was replaced by selectable marker, cells was then introduced a temperature-inducible copy of the cre site-specific recombinases, which enables efficient removal of Cm resistance cassette from chromosome, generating the strain iHKU101. The replacement and removal was confirmed by PCR (Fig.1B).

Fig. 1 Diagram illustrating the strategy used to delete cheZ (A) and agarose gel electrophoresis of confirmation PCR products (B). B, 1-4, MA6 (cheZ replaced with cm); 5, iHKU101; 6, MG1655.

Thereafter, we applied cells at log phase onto soft agar plate to test its motility. Cells lacking cheZ tumble incessantly and are essentially nonmotile, the spot expanded by only 0.3 cm (Fig2.B), while the parental strain, MG1655, has propagated across the plate (Fig2.A).

Fig.2 Migration of strains MG1655 and iHKU101 18h after seeding.
A, MG1655; B, iHKU101

[Back to top]


Introduction of cheZ restored the motility of strain iHKU101
To answer the question whether or not introduction of cheZ gene back to iHKU101 could restore its motility, we introduced cheZ gene under the control of plac/ara-1, whose transcription is induced by either IPTG or arabinose. SDS-PAGE and western blotting analysis showed the high expression level of cheZ in the presence of IPTG or arabinose (Fig.3A). And arabinose exerted a higher inductivity. We also observed a moderate amount of CheZ expression in absence of any inducers, indicating the leaky of plac/ara-1. This might due to the lack of lacIq in the chromosome of iHKU106.
We then plated the iHKU106 cells onto soft agar plate containing arabinose to test its motility. As shown in Fig.3B, iHKU106 migrated about 4 cm (diameter) from the center of the plate after 10 h incubation at 37C.

Fig.3 SDS-PAGE and western blotting analysis of CheZ expression (A) and migration of strains iHKU106 (B).

[Back to top]


Pattern formation

We confirmed motility is restorable (or inducible) in E. coli. We proceeded to investigate whether a unique pattern can be obtained by our density-dependent motility design. We constructed the 2 strains, the high-density mover and the low-density mover, as described in the Design Section and let them, together with controls, to grow on soft-agar-plates following the “pattern development on soft-agar-plate” protocol. Specifically, the strains were pre-culture from overnight inoculums, when reach OD600 ~0.8, 2 μl of the culture was applied to the center of a soft-agar-plate and the plate was subsequently incubated overnight. We obtained quite different patterns amongst different strains.

    Circle pattern from MG1655 (wild type Control)

In the case of the wild type, the obtained pattern was a circle (Fig. 4A). As can be seen in the density profile (Fig. 4B), cells are uniformly distributed across the area covered. Similar patterns were generated by the model (Fig. 4CD) and known as Fisher wave; see the movie below Fig.4 for the dynamics of the expanding bacteria population.

Fig.4 Pattern of MG1655.
A, bacteria population on soft agar plate 12 hours after seeding at center; B, 3-D profile of A; C, pattern of wild type cell by model; D,3-D profile of C.

Supporting videos


    Mt.Fuji pattern from iHKU105

But in the case of the high-density mover iHKU105(design), a different pattern was observed compared to the wild type strain. In particular, the center of the circle is high in density (Fig. 5). Similarly, we got a pattern by our model, choosing an increasing function Dρ(h) (modeling).

Fig.5 Pattern of iHKU105.
A, soft-agar-plate; B, 3-D profile of A.

    Ring-like pattern
And in the case of iHKU102 (design), more distinctive pattern was observed compared to the wild type strain. In particular, a region of low cell density was observed (Fig. 6). We have termed the inner ring with the lowest density the ring of void and its diameter is roughly 1.3cm. To note that the diameter of this ring of void would not change once it formed.

In our model, if the Dρ(h) decreased smoothly near the threshold, an amazing identical ring-like pattern was observed (modeling).

Fig.6 Pattern of iHKU102.
A, soft-agar-plate; B, 3-D profile of A; C, modeling pattern; D,3-D profile of C

Supporting videos


[Back to top]


    Effect of genetic circuit modification on pattern formation

As predicted by modeling, if the graded change of cheZ expression to changes in cell density in iHKU102 is turned into an abrupt response, the ring-like pattern formed by iHKU102 will be different, i.e. multiple rings are expected. This can be done by placing the expression of LuxI, LuxR, and the target proteins all under the control of pluxI as demonstrated by Haseltine et al.[2]. So in iHKU117, pluxRI2 was replaced with pluxRI3, by which luxR and luxI are controlled by positive auto-feed back. Interestingly, a different pattern really appeared. It’s apparent that the inner area shrank, and there’s a circle of cells with high density surrounding a ring of void. Outside this circle of cells, there’s the other weaker ring of void (Fig. 7).

Fig.7 Patterns of iHKU102 and iHKU117.
A, soft-agar-plate; B, 3-D profile of A; C, modeling pattern of iHKU117; D,3-D profile of C.

[Back to top]

    Multiple initial spot
We tested patterns resulting from a variety of different initial seeding, e.g., two spots at different distances and 3 spots. As shown in Fig. 8, the patterns obtained (left panel) are in good agreement with the model prediction (right panel) in all cases.


Fig8. Multiple initial spot with two or more initial droplets in modelling(left)and in experiment(right)

Importance of genetic circuits in pattern formation

To determine the importance of introduced genetic circuit in this ring-like pattern, a series of control strains were generated (Plasmids and strains). As shown in Fig.9, loss of any of the following elements -- luxR (iHKU104),luxI (iHKU103), CI (iHKU109), luxR and luxI (iHKU115), or CI and cheZ (iHKU118)-- resulted in the failure of development of ring-like pattern (iHKU102), suggesting that a complete genetic circuitry is required.

Fig. 9 photograph of control strains

[Back to top]

Characterization of factors controlling ring-like pattern formation

    Effects of temperature

We cultured the strains at room temperature (23oC), 32oC, and 37oC to investigate the effects of temperature on pattern development. Except for the time to form the ring-like pattern, the overall pattern appeared to be unaffected by temperature (data not shown).

    Effects of humidity

During our experiments, we sought to determine the effects of humidity of the environments on the pattern formation. When the humidity is too low, the water in the agar plate will be evaporated into the air rapidly, the agar concentration will rise, which will then cause the change of the diffusion coefficient of cells.
In order to control the humidity in a pattern formation favored range, we did some experiments to try to control it. We used a big file box with 14 agar plates in it in the 37 centigrade warm room, which is the normal experiment condition. At the same time, plates with water (100 ml per plate) were put into the box too. By changing the number of them, we attempted to control the humidity in the box at different level. Each time, there were a fixed number of plates of water in the box, after sufficient time, the humidity should be in equilibrium. We expected to have different levels of humidity for different number of plates of water.
We found that the change of number of water plates do change the humidity in the box, however, there is only 10% increase in humidity (from 73% to 83%) in the presence of 4 water plates. At the same time, to our surprise, we found that if the box was sealed tightly, even without water plates in it, the humidity will reach ~80 %. We argue that this may be due to water evaporation of the agar plates.

Fig.10 effects of water amount on humidity

    Effects of agar hardness

During our experiments, we found that the agar concentration of the soft-agar plates is an influential factor to the development of the pattern. We investigated this factor by growing our strain on plates over a range of agar concentration. As shown in figure 11, the ring of void becomes clearer and clearer as agar concentration increases. This sheds light on cell diffusion, since the denser the agar, the slower the cell diffusion through the matrix.

Fig.11 effects of agar concentration on ring-like pattern

[Back to top]


Quantitative measurements

    Grwoth rate in batch culture

    Growth curve of iHKU101

iHKU101 is measured as a control for the experimental strain. After two times of precultures, the initial cell concentration was diluted to be OD600~0.05, and it entered exponential phase after two doublings and started to go into stationary phase after 100 minutes at OD600~1.6.
The doubling time is 20.03min (fig.12), which is shorter than that of iHKU102 (below).

Fig. 12 Growth curve of iHKU101 with the doubling time ~ 20.0 min

    Growth curve of iHKU102

The doubling time of iHKU102 is generally longer than iHKU101 (more than 30min) even without antibiotics (Fig.13). The initial OD is also ~0.05 and it enters exponential phase immediately in the experimental culture.

Fig. 13 Growth curve of iHKU102.

A, The doubling time ~ 33.6 min in absence of antibiotics, and B, ~36.5 min in the presence of antibiotics.

    Growth rate on agar plate

    Measured by total protein assay

In the measurement of the growth on plates, the total protein within the agar, which is an index of the E.coli population, was measured by (Bradford method) for a time range of 12 hours with an interval of one hour.
We want to find out how the cell mass develops on the agar plate over the time. Unlike the LB batch culture, which can be measured by OD600 directly, to know the cell density on soft agar plate requires special measurements. One is total protein and the other is the optical measurement. Since it is difficult to harvest sample of a chosen mass/volume, we harvest the whole agar from the 35mm petri-dish, which is of the exact volume of 1 ml.
Since the cells are mixed evenly in the agar, it doesn’t resemble the situation on plate for pattern formation. However, the similarity of the trend of the curves from the optical methods and this total protein method reassures the validity of the optical methods in the determination of cell population on plate. What’s more, it provides information about the growth curve of specific strains/conditions, which is important in the modeling.

Fig.14 Growth curve of iHKU102 on soft agar plate (total protein).

The total protein at different time points was measured, which were converted to OD600 values (Fig.14), according to a formula generated before (data not shown). The first time point is before the cells were added to the agar, and put on ice. However the reading is 0.8944, rather than zero, which may be due to the protein in the LB agar. Therefore it is needed to subtract the blank from the following readings when estimating the cell population on the agar. Generally the curve increases over time, with one outliner of the third one, which also appears in the brightness result. The total protein level reaches the maximal at 10 h, with a value about twice as much as the initial protein level, and then begins to drop.

    Measured by brightness

Another method to measure cell growth on plate is by analysis of the brightness. This method can measure the cell density at any specific location of the plate without the need to harvest sample. However there is a drawback for this method -- variance may be due to the uneven distribution of the light reflected from the plate.
To investigate the relation between brightness and cell density, we captured the images of plates with given ?cell density by using our novel device (movie taker). By calculating the brightness of each plate, we created a graph with OD600 as X-axis and brightness intensity as Y-axis, and obtained a formula of the trend line: y = 24.229ln(x) + 48.962 (Fig.15). Thus we can convert the brightness intensity value into the OD600 value.

Fig.15 Relation between brightness intensity and OD value.

A, photograph of plates containing inceasing cell density; B, trend line and formula indicating the relation between brightness intensity reflected from plates in panel A and OD600 value.

We attempted to test if these two methods were comparable or not. So the optical analysis was conducted simultaneously with the total protein analysis at each time point. For each time point, we took a picture of the plate, and then analyzed the brightness at each pixel and added up for the total light intensity of the whole plate. Finally the brightness intensity was converted into an OD600 value and a growth curve was generated (Fig.16).

Fig.16 Growth curve of iHKU102 on soft agar plate (brightness).

We next sought to measure the growth rate of iHKU102 on ring-like pattern plate. Since the total protein method is difficult to conduct in this case, we measure the brightness intensity of different region on ring-like pattern plate. In figure 17B, the brightness intensity of inner area increased in the first 15 h and then stayed at the peak in following hours. The outer area appeared at 9 h and increased gradually over time (Fig. 17C).

Fig.17 Growth curve of iHKU102 on ring-like pattern plate.
A, photograph of the ring-like pattern, the inner and outer area are indicated; B, growth curve of cells within the inner area; C, the outer area; D, both areas.

    Front propagation speed on agar plate

By using automatic brightness measurement, Front propagation speed of iHKU102 was measured in the first 15 h. The Front propagation speed can be well deducted by measuring the diameters of the inner and outer area edge. The Front propagation speed of iHKU102 on 0.45% agar plate was 1.33 mm/h (Fig.18A). We next measured the Front propagation speed of iHKU102 cells in agar with lower concentration (Effects of agar hardness), and found that the Front propagation speed in 0.45 % agar (Difco) plate is 1.33 mm/h. It is much lower than that of 0.35 % agar (Difco) plate (1.53 mm/h), in which no clear ring was observed (Fig.18B; Fig.11). This data supports our hypothesis that Front propagation speed significantly affects the development of ring-like pattern.

Fig.18 Migration rate of iHKU102 on ring-like pattern plate

    Levels of LacZ and corresponding growth curves.

The expression of cheZ was measured by LacZ assay. By replacing pRg in iHKU105 with pRglacZ, we generated the strain iHKU114. In this strain, the regulatory genes luxI (autoinducer synthase gene) and luxR (autoinducer receptor protein gene) are controlled by plac/area-1 and supposed to constitutively express in the presence of arabinose or IPTG; the expression of lacZ gene are driven by pluxI. Consequently, iHKU114 cells were expected to induce β-galactosidase synthesis in a cell density-dependent fashion. This was found to be the case (Fig.19A). lacZ expression was studies along with the growth curve of strains (Fig.19B). The measured LacZ activity was corrected by normalized to protein concentration and expressed in fluorescence per mg. When iHKU114 cells were inoculated into fresh medium, β-galactosidase activity decreased at the early 40 min, rose to a peak. It returned to the initial level at 70 min, and remained constant for 1.5~2 h during growth and was then induced to a level approximately 3-fold higher than preinduction levels in following 5 h. These results are consistent with the data reported previously [3], indicating a dependence on autoinducer for pluxI transcription.
Since the endogenous lacZ gene was not deleted from the chromosome of iHKU114, to rule out its influence, we included the parental strain iHKU101 as negative control and used arabinose as inducer. The result showed that no apparent β-galactosidase activity was observed in iHKU101 (Fig.19A).

Fig. 19 Levels of LacZ (panel A) and corresponding growth curves (panel B) of iHKU114 and iHKU101.

To study cheZ expression in iHKU102, we replaced cheZ gene with lacZ gene and obtained the strain iHKU108. In this strain, lacZ expression is under the control of λpR-O12, which is repressed by a strong λ repressor, CI. And CI is driven by pluxI. Thus it was anticipated that λpR-O12 would be repressed in a cell density-dependent manner, and thereby lead to a decrease in the levels of β-Galactosidase activity.
β-Galactosidase activity in this strain was substantially lower than in iHKU114. We reasoned this as the effect of the basal level expression of CI. When iHKU108 cells were inoculated into fresh medium, the β-galactosidase activity was at a low level, and rose to a constant level in 1.5 h. β-Galactosidase remained constant for 1.5 h, and then, as expected, decreased to near basal level in 0.5 h (Fig.20A). We observed that the β-Galactosidase activity levels of both iHKU114 and iHKU108 were induced/repressed when the cell densities reach OD600~2 (Fig.19A and 20A). This confirmed the cell density-dependent manner of our design.

Fig. 20 Level of LacZ (panel A) and corresponding growth curve (panel B) of iHKU108.

[Back to top]



To investigate the development of the obtained patterns, a movie documenting the duration of the pattern development would bring us invaluable information to look into the mechanisms. We therefore constructed a set of photo capturing devices to record snapshots of the plates during the course of their development (movie taker).


(Upper panel, experimental videos of MG1655, lower panel modeling videos)

The original length of this movie is 10 hours 30 minutes, and it is compressed in to 14 seconds. The compression ratio is 2700:1. Similarly, This movie starts when cells can be clearly seen on the plate, i.e. cells growth after our dropping cells on the plate until the start of this movie is omitted, and this omitted part corresponded to 1 hour. Compared to the previous movie, no such clear pattern can be observed during the whole process. Bacteria migrates outwards normally till the end of the movie. A migration speed of the bacteria can also be measured. Please refer to table.1 for detailed information.

[Back to Fig.4]



(Upper panel, experimental videos of iHKU102, lower panel, modeling videos)

The original length of this movie is 11 hours, and it is compressed in to 16 seconds. The compression ratio is 2475:1. This movie starts when cells can be clearly seen on the plate, i.e. cells grows after dropping cells on the plate until the start of this movie is omitted, and this omitted part corresponded to 1 hour. According to the movie, it tooks 5 hours for the bacteria to form the inner ring and the ring ablized at the very location. Another 2.5 hours was needed for a vague outer field to be seen. Bacteria in the outer part will continue to grow as well as migrate outwards normally after the end of this movie. By applying the light intensity measurement the cell number inside the void ring is evaluated. Amigration speed of the bacteria can also be measured. The cell concentration inside the void ring is measured by our brightness method. Please refer to the table below for detailed information.

[Back to Fig.6]









    1. Datta S, Costantino N, Court DL. A set of recombineering plasmids for gram-negative bacteria. Gene. 2006, 379: 109-15.
    2. Haseltine EL, Arnold FH. Implications of rewiring bacterial quorum sensing. Appl Environ Microbiol. 2008, 74: 437-45.
    3. Dunlap PV, Kuo A. Cell density-dependent modulation of the Vibrio fischeri luminescence system in the absence of autoinducer and LuxR protein. J Bacteriol. 1992, 174: 2440-8.