Logo of emborepLink to Publisher's site
EMBO Rep. 2009 May; 10(5): 527–532.
Published online 2009 Mar 27. doi:  10.1038/embor.2009.25
PMCID: PMC2680869
Scientific Report

Direct observation of type 1 fimbrial switching


The defining feature of bacterial phase variation is a stochastic ‘all-or-nothing' switching in gene expression. However, direct observations of these rare switching events have so far been lacking, obscuring possible correlations between switching events themselves, and between switching and other cellular events, such as division and DNA replication. We monitored the phase variation of type 1 fimbriae in individual Escherichia coli in real time and simultaneously tracked the chromosome replication process. We observed distinctive patterns of fim (fimbriae) expression in multiple genealogically related lineages. These patterns could be explained by a model that combines a single switching event with chromosomal fim replication, as well as the epigenetic inheritance of expressed fim protein and RNA, and their dilution by growth. Analysis of the moment of switching at sub-cell-cycle resolution revealed a correlation between fim switching and cell age, which challenges the traditional idea of phase variation as a random Poissonian phenomenon.

Keywords: fim, fimbriae, SeqA, single cell, phase variation


Phase variation is characterized by a stochastic switching between discrete gene expression states (Henderson et al, 1999; van der Woude & Baumler, 2004). This form of gene regulation contrasts with classical gene regulation, both in terms of the underlying molecular mechanisms and the effect on the population structure. Switching between phases typically involves a genetic or epigenetic mechanism such as a DNA inversion or a loss or gain of DNA methylation, resulting in ‘all-or-nothing' ON or OFF expression levels. As the switch state is heritable over multiple generations, it allows the population to differentiate and stably maintain sub-populations. Phase variant macromolecules are typically exposed on the cell surface, and have a crucial function in bacterial virulence and immune evasion.

The regulation of phase variation switching rates has been studied in detail for several systems (Blomfield, 2001). However, the experimental methods used so far do not capture the switching event itself, but rather infer switching rates by monitoring the ratio between ON and OFF cells within a population (Gally et al, 1993). For this reason, the most elementary questions on the correlations and dynamics of phase variation have remained unaddressed. For example: is there a correlation between switching events in genealogically related cells? Does switching occur in particular phases of the cell cycle? What are the expression dynamics during switching? Here, we investigate these issues by following the expression of a phase-variable gene in single cells and in real time. In particular, we investigate the switching behaviour of type 1 fimbrial expression in Escherichia coli.

The mechanism of fim (fimbriae) switching is characterized by a reversible DNA inversion event, in which a cis regulatory element that contains the promoter of the fim operon changes its orientation (Fig 1A). The DNA inversion process involves the formation of a DNA recombinase (FimB and/or FimE) complex forming a loop of the 314-bp invertible region, the subsequent breaking and rejoining of single DNA strands, and the dissipation of the synapse complex to allow transcription of the fim operon.

Figure 1
The fim operon in ON and OFF phases. (A) In the ON orientation, the promoter positioned within the invertible element fimS drives transcription of the fim operon. GFP has been inserted into fimA, the first gene of the fim operon, allowing single-cell ...

To detect fim switching events in single cells, GFPmut2 was inserted into the chromosomal fimA, which is the first gene of the fim operon. We followed cell growth and the level of green fluorescent protein (GFP) expression over time, and determined the genealogical relationships within the population by using image analysis. Chromosome replication was simultaneously monitored in some experiments, using a fusion between the DNA sequestering protein SeqA and the mCherry fluorescent reporter. SeqA forms visible foci at replication forks, as it has a binding affinity for hemimethylated DNA (Hiraga et al, 2000; Yamazoe et al, 2005).


Phase variation in real time at the single-cell level

To characterize the distribution of the fim operon expression level at a single time point, isogenic cells growing exponentially in defined rich medium were spread onto agar and then imaged using phase-contrast and fluorescence microscopy. Two distinct sub-populations—one with a low mean GFP brightness per unit area (0.77±0.23a.u.) and another with a high brightness (8.18±1.88a.u.; Fig 1B)—corresponding to the OFF and ON states of fimS, were observed. The ON cell fraction (4.6%) was in agreement with previous studies (Leathart & Gally, 1998). The low level of expression in the OFF state, which is indistinguishable from cells lacking the GFP fusion, was consistent with the tight orientational specificity of the promoter.

Cells were monitored as they continued to grow and divide. Microcolonies of about 500 cells were formed over a period of approximately 5 h (generation time, ~25 min; supplementary Fig S1 online). Some descendants of a non-fluorescent OFF progenitor spontaneously showed an increased fluorescence that signalled ON switching, resulting in a heterogeneous colony with patches of different phases (Fig 2A; supplementary Movie S1 and S2 online). Conversely, brightly fluorescent ON progenitors produced descendants with a decreasing fluorescence, indicating OFF switching (supplementary Movie S3 and S4 online). Cells grew exponentially throughout the experiment with a constant growth rate (supplementary Fig S1 online). In total, more than 400 progenitors were monitored producing more than 20,000 single-cell trajectories (see supplementary Table S1 online).

Figure 2
ON switching in a growing microcolony. (A) Overlay of fluorescence and phase-contrast images showing a single OFF cell developing into a microcolony. A cell in this isogenic population starts to fluoresce spontaneously (white arrow), and develops into ...

Distinct patterns of fim expression near switching

We constructed family trees detailing the genealogical relationships within the microcolonies, as well as the GFP expression history of each lineage (Fig 2). Some lineages showed a monotonic increase in brightness, whereas other lineages showed an increase followed by a decrease. These expression profiles developed over the course of several generations, which is significantly longer than the GFP expression and maturation time of about 6 min (supplementary Fig S2 online). The observed decreases in brightness might originate from ON-to-OFF switching events. However, as the bulk ON-to-OFF switching rate (of the order 0.1 per cell per generation; Gally et al, 1993) is too low to explain the data, one would have to invoke as-of-yet unknown correlations between switching events. Instead, we hypothesize that the observed pattern of expression is caused by the chromosome replication process in combination with a single OFF-to-ON switching event.

To synthesize a sufficient number of chromosomes for all progeny, rapidly dividing bacteria are thought to have multiple nested replication forks that operate simultaneously (Yoshikawa et al, 1964; Cooper & Helmstetter, 1968). Consequently, genes positioned near the chromosome origin (oriC) are amplified in number, whereas those near the terminus (terC) are not. Within the E. coli chromosome, the fim system is located near oriC (at ~610,000 bp from oriC or ~13% of the whole chromosome), and should therefore be present in multiple copies.

How could multiple chromosomal copies of fim explain the observed pattern of expression? We consider a single OFF-to-ON event, in which initially all fimS copies are in the OFF state, of which one turns ON. In Fig 2C, this moment could coincide with the first fluorescence increase seen in the orange lineage at about 100 min. On division, this single chromosomal ON copy is inherited by one of the daughters (red lineage), which continues GFP expression. By contrast, the second daughter (orange lineage) inherits several fimS copies in the OFF state, as well as GFP proteins and messenger RNA, which will eventually be diluted due to growth resulting in a slow decrease in mean fluorescence. After the replication of fimS and division in the red lineage, one daughter (red) decreases in brightness, whereas the other daughter (green) continues to increase, suggesting that the latter inherited both ON copies and the former only OFF copies. This grouped redistribution of fimS copies agrees with a nested arrangement of replication forks, in which the most recently replicated DNA remains physically linked with the formerly replicated DNA. No other lineages with decreases were observed, suggesting that the green lineage inherited only ON fim copies. An inheritance of two fimS copies at birth is in accordance with the Helmstetter–Cooper model of chromosome replication at these growth rates (Cooper & Helmstetter, 1968). This pattern has parallels with the idea of gene dosage, which has been used to denote differences in the expression level owing to different copy numbers of the corresponding gene, for example due to gene duplication.

The model described above provides two predictions. First, the pattern of GFP expression should alter for different growth rates, as the chromosome replication scheme is expected to change with growth rate. Slow-growing cells with generation times greater than 60 min are expected to have one chromosome copy at birth and two at division (Cooper & Helmstetter, 1968). This would imply that one lineage at maximum would show one expression peak, instead of the two peaks observed for fast growing cells. Second, the pattern of expression should depend on the moment of switching within the cell cycle (Fig 3). In slow-growing cells, if switching occurs after birth but before replication of the fimS switch, then all descendants would show an increase in monotonic fluorescence. If, conversely, switching occurs after the replication of fimS, only one of the two fimS copies is ON at division, resulting in one daughter lineage with an expression peak.

Figure 3
Fimbrial expression and DNA replication. Model for slow-growing Escherichia coli cells inheriting a single chromosome (replicated in a bi-directional manner) with an OFF fim switch. At a certain moment, the switch spontaneously switches to the ON state ...

Dependence on growth rate and cell age

To test the predictions outlined above, cells were grown on succinate minimal medium with a mean generation time of approximately 120 min (fraction of ON cells, 7.4%). The number of SeqA foci within one cell cycle showed a regular 0-1-2-1-0 pattern (Fig 4E,F). The appearance of a single SeqA focus indicated the start of chromosome replication (Fig 4G). The single focus then split into two separate foci, indicating a bi-directional movement of the two replication forks, which later merged again into one focus, suggesting that the forks were near termination. The disappearance of the focus indicated the complete methylation of both chromosomes and the end of replication. The delay in methylation of nascent DNA has been reported to be about 1.5 min (Lu et al, 1994). The moment of fimS replication was determined at 26% of the replication process, based on a bi-directional replication and a constant replication rate along the chromosome. The moment of fimS switching was determined by the intersection of the background fluorescence signal and the rising fluorescence signal, while correcting for the GFP maturation time.

Figure 4
ON switching events in slow-growing cells. (A) OFF-to-ON switch with transient expression increase. (B) OFF-to-ON switch with only increases in monotonic expression. (A,B) Top panels show phase-contrast images overlaid with GFP fluorescence signal reporting ...

In this manner, we were able to determine the number of inherited chromosomes, to pinpoint the start and end of replication (and consequently the moment of fimS replication), and to correlate it with fim switching time and the ensuing pattern of GFP expression of individual cells. The data showed that all fimS switching events occurring after the replication of fimS resulted in a single lineage with transient GFP brightness (Fig 4A,C,E). All fimS switching events occurring before the replication of fimS resulted in no lineages with transient brightness (Fig 4B,D,F). The results at slow growth verify the essential feature of our proposed model: the pattern of expression is determined by the growth rate and switching moment relative to the moment of fimS replication. For ON-to-OFF switching, a similar but inverse pattern of GFP expression was observed (supplementary Fig S3 online), which showed decreases in brightness followed by increases (back to the ON fluorescence level), as well as monotonic decreases in the fluorescence level.

Other details of the expression traces could also be understood within the framework of the replication model. For example, as switching before replication quickly results in two fimS copies in the ON state, the expression rises faster, leading to a more rapidly established steady-state expression level (corresponding to 100 min for the first case compared with ~200 min for the second case).

Higher fim switching rate at the beginning of cell cycle

Next, we investigated how the fimS switching events are distributed within the cell cycle. The null hypothesis we aimed to test is that for one given environmental condition, fim switching occurs randomly in time with a constant switching probability per fimS throughout the cell cycle. We found that replication of fimS occurs at an average cell age of ϕ=0.36±0.11 (average value of each cell cycle's B-period plus 0.26 of its corresponding C-period, divided by its Td). Here, cell age denotes the progression within the cell cycle—running from 0 at birth to 1 at division—and is thus distinguished from the cell-pole age that can run over multiple divisions (Stewart et al, 2005). Given the doubling of fimS copy number on replication and the null hypothesis, the expected fraction of switching events that occur before the replication of fimS would be ϕ/(ϕ+2(1−ϕ))=0.22. Our data show that the measured fraction is 0.44 (24 out of 55 OFF-to-ON switching events in slow-growing cells), which is significantly higher than expected (χ2 test, P<0.001). The distribution of switching probability in time (Fig 4H) shows that the switching probability is approximately two times higher in the first part of the cell cycle.


By following single cells in real time, we have shown not only a distinctive pattern of expression of the fim operon on fim switching by DNA inversion, but also that they are explained by accounting for the number of chromosomes and the moment of fim switching with respect to the moment of replication of the fim switch. The observed patterns are consistent with independent switching of each fim copy, as expected when local events at the fim switch, such as DNA looping or the actual inversion reaction, are the origin of the switching stochasticity. In a possible alternative pattern, a fluctuating global cellular parameter such as the concentration of recombinase is the origin of the stochasticity. In this case, switching would occur only after the recombinase concentration exceeds a critical value. Consequently, the multiple fim copies in one cell would then switch in a time-correlated manner. However, the data are not consistent with such a pattern. We did not observe switching events after fim replication in which both daughters showed monotonically increasing fluorescence. In addition, it has previously been found that when recombinase FimB is overexpressed, switching remains a rare stochastic process (~10−2 events per cell per generation; Dove & Dorman, 1996), which also suggests that processes other than fluctuations in recombinase expression level are the causes of the large switching timescales.

Unlike other phase variation control mechanisms such as slipped-strand mispairing or specific Dam methylation sites, the inversion of fimS is not expected to require DNA replication as an essential step in the switching process (Low et al, 2001; Davidsen & Tonjum, 2006). However, a crucial finding of this study is that OFF-to-ON fim switching occurs preferentially at the beginning of the cell cycle, before replication of the fim genes. What might be the cause of this observed switching bias? The known molecular mechanisms underlying fimS switching do not provide a convincing answer, and we might only speculate on possible causes. One might consider variations in the level of recombinase fimB (required for OFF-to-ON switching). The expression of fimB is diminished considerably in a dam mutant (supplementary Fig S4 online), and the generation of unmethylated—or hemimethylated—DNA in the vicinity of fimB following passage of the replication fork might temporarily reduce fimB expression. Although this pattern would suggest a more active and regulated link between cell cycle and gene regulation, at present we do not understand how Dam affects fimB expression. However, we have ruled out the possibility that this is due to changes in methylation protection associated with the interaction of activator proteins, NanR and NagC, with their operator sites upstream from fimB (Sohanpal et al, 2004; data not shown).

Alternatively, the observed bias for switching at the beginning of the cell cycle might be caused by changes in DNA topology during replication. The site-specific recombination process that is at the basis of fimS switching has been shown to be sensitive to changes in DNA topology (Dove & Dorman, 1994). During DNA replication, the separation of complementary DNA strands by helicase results in a relaxed (or possibly positive supercoiled) region in front of the replication fork, and a negative supercoiled region behind it (Kornberg & Baker, 1992). The fim switching rates might thus be modulated by these changes in supercoiling density or, instead, by the disruption of the recombinase synapse with the looped fimS by the passing replication fork. A possible advantage of switching early in the cell cycle might be that it reduces the number of transiently expressing, ambiguous lineages, thus limiting the wastage of resources on unused fim protein and RNA.

By investigating the behaviour of single cells, this study has shown a correlation between the cell-cycle dynamics and fim switching behaviour. It shows that multiple genetic fim systems per cell ought to be taken into account in mechanistic descriptions of fim switching. The observed modulation of the fim switching rate with the cell cycle challenges the common idea that fim switching is a straightforward Poissonian process, in which the switching probability is constant in time for fixed external conditions. The methodologies presented here can be applied to explore further the boundary between randomness and correlations in phase variation, as well as to reveal the dynamics of other phase variation systems at the level of discrete events.


Strains and plasmids. Derivatives of E. coli K-12, strain ASC129 (MG1655 fimAΩGFPmut2) and strain ASC215, which is ASC129 bearing pASC215 (AmpR, pBR322 ori; rop, mcherry-seqA), were used for experiments with and without SeqA foci. For microscopy experiments, EZ Defined Rich medium and MOPS medium (Teknova, Hollister, CA, USA) supplemented with 0.4% sodium succinate (Sigma Aldrich, St Louis, MO, USA) were used. Ampicillin 100 μg/ml (Sigma Aldrich) was added to the medium when appropriate.

Microscopy sample preparation. A fresh overnight culture was diluted with pre-warmed medium (such that the optical density of the culture would never exceed 0.02) and grown for a further 5 h to be in the exponential phase.

Two microscope slides, one with a hole of size 18 mm × 55 mm, were attached to each other by using silicon grease (Dow Corning, Midland, MI, USA) and heated to 80 °C. Next, 10 × concentrated medium was mixed with hot 2% agaroseMP solution to its final concentration (total volume is 1 ml) and spread horizontally across the centre of the cavity of the microscope slides. Immediately afterwards, the slides were transferred to a cool flat surface, and covered with a 24 mm × 60 mm silanized coverslip (a normal coverslip dipped into RepelSilane and then left to dry). The coverslip was placed directly on top of the liquid agar so that its surface would become flat on cooling (at room temperature).

Once the agar had solidified, the slides were moved to the microscope's 37 °C incubation chamber (to avoid any cold shock to the cells when they were deposited on the agar surface). Inside the chamber, the coverslip was removed. The now flat-surfaced agar was trimmed into a thin strip of about 2 mm × 55 mm across the centre of the cavity. The rest of the cavity acted as an air reservoir for cell growth. By using a pre-warmed pipette tip, 2–3 tiny droplets of the culture were dropped on the agar strip. Then, the sample was closed immediately with a normal pre-warmed coverslip and sealed with silicon grease. The cells spread evenly and became immobile. Over time, each single cell grew into microcolonies.

Microscopy and data analysis. Imaging was performed with a Nikon Eclipse TE2000 inverted microscope. Cell fluorescence quantification and lineage tree construction were performed with Schnitzcell, a program written in MATLAB, kindly provided by M. Elowitz. A further description of Materials and methods is provided in the supplementary information online.

Supplementary information is available at EMBO reports online (http://www.emboreports.org)

Supplementary Material

Supplementary Movie S1

Supplementary Movie S2

Supplementary Movie S3

Supplementary Movie S4

Supplementary Information


This study on the dam mutant (supplementary Fig S4 online) was performed by B.K. Sohanpal. We thank T. Brendler and S. Alexeeva for the plasmids; E.J. Stewart, D. Gadella and N. Vischer for help in early microscopy work; R. Tsien for mCherry; and C. Woldringh and T. den Blaauwen for useful discussions. We also thank M. Elowitz for providing the Schitzcell program and R. Dries for technical support. This study was supported by the organization for Fundamental Research on Matter (FOM), which is financially supported by the Netherlands Organization for Scientific Research (NWO). Work in the I.C.B. laboratory was supported by grant 076360/Z/05/Z from the Wellcome Trust.


The authors declare that they have no conflict of interest.


  • Blomfield I (2001) The regulation of pap and type 1 fimbriation in Escherichia coli. Adv Microb Physiol 45: 1–49 [PubMed]
  • Cooper S, Helmstetter CE (1968) Chromosome replication and the division cycle of Escherichia coli B/r. J Mol Biol 31: 519–540 [PubMed]
  • Davidsen T, Tonjum T (2006) Meningococcal genome dynamics. Nat Rev Microbiol 4: 11–22 [PubMed]
  • Dove SL, Dorman CJ (1994) The site-specific recombination system regulating expression of the type 1 fimbrial subunit gene of Escherichia coli is sensitive to changes in DNA supercoiling. Mol Microbiol 14: 975–988 [PubMed]
  • Dove SL, Dorman CJ (1996) Multicopy fimB gene expression in Escherichia coli: binding to inverted repeats in vivo, effect on fimA gene transcription and DNA inversion. Mol Microbiol 21: 1161–1173 [PubMed]
  • Gally DL, Bogan JA, Eisenstein BI, Blomfield IC (1993) Environmental regulation of the fim switch controlling type 1 fimbrial phase variation in Escherichia coli K-12: effects of temperature and media. J Bacteriol 175: 6186–6193 [PMC free article] [PubMed]
  • Henderson IR, Owen P, Nataro JP (1999) Molecular switches—the ON and OFF of bacterial phase variation. Mol Microbiol 33: 919–932 [PubMed]
  • Hiraga S, Ichinose C, Onogi T, Niki H, Yamazoe M (2000) Bidirectional migration of SeqA-bound hemimethylated DNA clusters and pairing of oriC copies in Escherichia coli. Genes Cells 5: 327–341 [PubMed]
  • Kornberg A, Baker J (1992) DNA Replication. New York, NY, USA: W.H. Freeman & Co
  • Leathart JB, Gally DL (1998) Regulation of type 1 fimbrial expression in uropathogenic Escherichia coli: heterogeneity of expression through sequence changes in the fim switch region. Mol Microbiol 28: 371–381 [PubMed]
  • Low DA, Weyand NJ, Mahan MJ (2001) Roles of DNA adenine methylation in regulating bacterial gene expression and virulence. Infect Immun 69: 7197–7204 [PMC free article] [PubMed]
  • Lu M, Campbell JL, Boye E, Kleckner N (1994) SeqA: a negative modulator of replication initiation in E. coli. Cell 77: 413–426 [PubMed]
  • Sohanpal BK, El-Labany S, Lahooti M, Plumbridge JA, Blomfield IC (2004) Integrated regulatory responses of fimB to N-acetylneuraminic (sialic) acid and GlcNAc in Escherichia coli K-12. Proc Natl Acad Sci USA 101: 16322–16327 [PMC free article] [PubMed]
  • Stewart EJ, Madden R, Paul G, Taddei F (2005) Aging and death in an organism that reproduces by morphologically symmetric division. PLoS Biol 3: e45. [PMC free article] [PubMed]
  • van der Woude MW, Baumler AJ (2004) Phase and antigenic variation in bacteria. Clin Microbiol Rev 17: 581–611 [PMC free article] [PubMed]
  • Yamazoe M, Adachi S, Kanaya S, Ohsumi K, Hiraga S (2005) Sequential binding of SeqA protein to nascent DNA segments at replication forks in synchronized cultures of Escherichia coli. Mol Microbiol 55: 289–298 [PubMed]
  • Yoshikawa H, O'Sullivan A, Sueoka N (1964) Sequential replication of the Bacillus subtilis chromosome. 3. Regulation of initiation. Proc Natl Acad Sci USA 52: 973–980 [PMC free article] [PubMed]

Articles from EMBO Reports are provided here courtesy of The European Molecular Biology Organization
PubReader format: click here to try


Related citations in PubMed

See reviews...See all...

Cited by other articles in PMC

See all...


  • Gene
    Gene links
  • GEO Profiles
    GEO Profiles
    Related GEO records
  • MedGen
    Related information in MedGen
  • PubMed
    PubMed citations for these articles
  • Substance
    PubChem Substance links

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...