Many bacterial species produce metabolites that accumulate in the extracellular environment and induce specific transcriptional responses in producing cells. This phenomenon, most often referred to as quorum sensing, is thought to constitute a self-cell-density sensing mechanism allowing bacterial populations to alter gene expression in response to increases in their own density. Quorum sensing systems involving N-acyl-l-homoserine lactone (AHL) production and response are the most intensively investigated example. In this study we have employed a novel technique, known as dielectrophoresis, to investigate the impact of colonial architecture on the induction of AHL mediated gene expression. Using dielectrophoresis, we constructed artificial mixed species microcolonies with specific architectures. In this way, we were able to show that approximately 1000 Escherichia coli cells layered over an immobilised cluster of approximately 500 AHL responsive cells alters the response of this cluster to AHLs supplied either exogenously or endogenously. These findings lend credence to the hypothesis that the accumulation of extracellular metabolites signifies generic crowding in mixed species assemblages.
A broad range of bacteria, including many species of environmental and medical significance, co-ordinate gene expression amongst cells through quorum sensing (QS) systems. QS is a phenomenon whereby extracellular accumulation of signalling metabolites enables bacteria to acquire information about neighbouring cells, and therefore constitutes one of life's most ancient forms of social behaviour . The most intensively studied QS system involves N-acyl-l-homoserine lactone (AHL) production and response. Over 40 bacterial species are known to encode AHL based QS systems . These species are scattered throughout the alpha, beta and gamma proteobacterial taxa and are derived from diverse environments . Many of the phenotypes regulated by AHL based QS are of medical, agricultural, evolutionary and ecological significance, including the production of virulence factors involved in human and plant disease, the conjugal transfer of plasmids and the generation of luminescence, surfactants, pigments and extracellular proteases .
For the most part, investigations into AHL mediated gene expression have used pure cultures to characterise the intracellular molecular mechanisms underlying the phenomenon. In contrast, there are very few studies that explore AHL mediated gene expression in the context of structured mixed species biofilm-like assemblages . Besides specific disruptive interactions with AHL receptor proteins  or lactonases , there are several conceivable mechanisms by which the extracellular concentration of AHLs could be altered by non-AHL producing cells in close proximity to AHL producing cells. For example, proximal non-AHL producing cells may adsorb or absorb AHLs or facilitate degradation through changes in environmental factors such as pH . Further, biotic crowding could easily affect other aspects of AHL mediated gene expression besides AHL accumulation, for example, through the depletion or generation of resources required for phenotype expression. If one assumes that the biotic architecture of a cell's local environment plays a significant role in the behaviour of that cell, then it is reasonable to revisit basic AHL mediated responses in a mixed lineage context.
To this end, we have developed electrokinetic protocols that enable the microscopic manipulation of colonial architecture such that mixed species assemblages with predetermined structure can be constructed. Pohl  pioneered the use of alternating currents to create non-uniform electric fields into which particles such as bacterial cells can be drawn. This electrokinetic process, commonly referred to as dielectrophoresis (DEP), was further developed by Alp et al.  and Verduzco-Luque et al. , enabling the construction of artificial microcolonies with defined structure. To date, however, cellular interactions within DEP constructed microcolonies have not been demonstrated. In this study, we have used DEP to create artificial microcolonies of AHL producing and AHL responsive cells. To these microcolonies, additional non-AHL producing cells were added, to test the null hypothesis that AHL mediated gene expression responses are necessarily identical in the presence or absence of biotic crowding from non-AHL producing cells.
2Materials and methods
2.1Strains and culture conditions
The AHL reporter strain used in this study was Escherichia coli JM105 harbouring the plasmid pJBA89, which encodes the luxR gene from Vibrio fischeri and a gfp gene regulated by the luxI promoter . This strain produces a short-lived green fluorescent protein (GFP) in response to a range of 3-oxo and unsubstituted AHLs. The AHL producing strain used in this study was Agrobacterium tumefaciens KYC6 harbouring the plasmid pCF218 . This strain overproduces 3O-C8-HSL. The strain used to generate covering layers of cells was E. coli JM109, which does not produce or respond to AHLs. Before administering cells to the dielectrophoretic chamber, they were cultured to stationary phase, shaking in Luria–Bertani broth. Cultures were incubated overnight (16 h), the E. coli strains at 37 °C and the A. tumefaciens strain at 28 °C. The plasmids pCF218 and pJBA89 were maintained by supplementing the relevant media with 50 μg ml−1 spectinomycin or ampicillin, respectively.
2.2DEP apparatus construction
Glass microscope slides (75 mm × 37 mm) were imprinted with a chromium electrode (200 nm thick) by photolithography [13–15]. The array of elements covered an area of 25 × 25 mm. The elements consisted of parallel bars with sidebars (castellations) measuring 30 × 30 μm and 30 μm apart. An open-ended chamber, adapted from Alp et al.  was built over the electrode-slide by placing strips of PVC tape along two edges of the electrode, upon which a coverslip was secured. This produced a chamber with a volume of approximately 110 μl and 200 μm in height. Media, cells, AHLs and flocculent were added to or removed from the chamber using a micropipette.
2.3DEP microcolony construction
Cells were washed three times and resuspended in deionised water, to an approximate cell density of 2 × 108 cells ml−1, before introduction to DEP chambers. A 20 V, 1.2 MHz alternating current (AC) was applied to the electrode, producing non-uniform (dielectric) fields between the elements. The aligned castellations generated discrete areas of high field strength into which cells were attracted. Two minutes were allowed for the cells to accumulate, then excess cells were removed by flushing 150 μl of deionised water though the chamber. The total number of cells on the slide was quantified by releasing the current at this point, sampling the chamber and counting by flow cytometry. To this end, suspensions were stained with SYTO 16 (Molecular Probes, USA) and enumerated on a FACSCaliber flow cytometer (Becton Dickenson, UK). The number of cells in each microcolony was calculated by dividing the total number of cells on the slide by the number of microcolonies. The number of cells drawn into each microcolony is dependent on potential energy wells formed by the areas of high electric field non-uniformity resulting from the application of the alternating current . Because the potential energy wells between castellations are uniform, variability in the number of cells in each microcolony is theoretically minimal.
To immobilise cells trapped in the alternating current, 40 μl of 0.005% (w/v) polyethylenimine (Acros Organics) was drawn through the chamber. Additional cells could then be added, trapped and immobilised allowing layered structures to be built up. These structures were stable when the current was discontinued allowing the introduction of media and 3O-C6-HSL (Sigma) to the chamber without disruption of the artificial microcolonies. GFP production was facilitated by replenishing the media within the chamber every 15–20 min.
2.4Microscopy, image capture and analysis
Images were captured using a Nikon E600 epifluorescence microscope fitted with a rear-mounted filter-wheel. The FITC (green) filter, with an excitation wavelength of 490 nm, was used to visualise GFP. Images were captured with a Digital Pixel camera, linked to a Power Macintosh G3. The shutter speed was 5 s for fluorescence microscopy and 50 ms for phase contrast. All images covered a field of view of 720 × 500 μm showing 35 DEP formed microcolonies.
Images were analysed with Image J freeware (http://rsb.info.nih.gov/ij/). Fixed sized regions of interest, approximately half the area of artificial cell clusters, were placed so as to measure a sample of pixels displaying fluorescence and avoiding reflections from the metal of the electrode. GFP production was measured as average pixel intensity for each region of interest. The relative fluorescence readings presented are an average of 105 readings from four separately prepared slides each. Significant differences in fluorescence output were determined with the student T test.
3.1Constructing artificial microcolonies with an AHL reporter strain
The methods of Alp et al.  and Verduzco-Luque et al.  were developed further to produce discrete microcolony-like aggregates of bacterial cells. The electrodes designed had directly opposite rather than offset castellations, thereby generating discrete high AC field regions. This novel micro-electrode design generated approximately 50,000 artificial microcolonies on a single slide, enabling mass replication of microscopic bacterial interactions. To compare AHL mediated responses in the presence and absence of non-AHL producing cells, we first had to construct microcolonies, represented diagrammatically in Fig. 1, from an AHL reporter strain that expresses GFP in the presence of the signalling metabolite.
DEP was used to form microscopic aggregates of an E. coli strain, harbouring the AHL monitor plasmid pJBA89, between microelectrodes immersed in growth media on a glass slide. A flocculent was used to immobilise the cells into artificial microcolonies before the AC current was removed. This procedure generated clusters containing approximately 500 cells each (Fig. 2). To ensure that the DEP and flocculent treatments had not killed or rendered the cells physiologically incapable of responding to AHL, the GFP response of the monitor strain in the microcolonies to various 3O-C6-HSL doses was tested. For each AHL concentration tested, artificial microcolonies were created on four replicate slides. To initiate the time course, the media in which the microcolonies were constructed was aspirated and regularly replaced with fresh media containing a given concentration of 3O-C6-HSL. Without this regular replenishment, only low levels of green fluorescence were observed. Using epifluorescence microscopy, green fluorescence was monitored every 15 min until the response reached a plateau.
GFP production in the artificial microcolonies was first observed after 20 min and reached a plateau after approximately 1 h. Fig. 3 shows the increase in green fluorescence over time in response to 0.5, 1, 2 and 5 μM 3O-C6-HSL. The minimum dose (0.5 μM) was found to be stimulatory, and the response was saturated by 2 μM 3O-C6-HSL. These data reveal that artificial microcolonies of an AHL monitor strain, constructed in this manner, are physiologically capable of responding to 3O-C6-HSL in a dose dependent manner.
3.2Non-AHL producing cells decrease the response of artificial microcolonies to an exogenous 3O-C6-HSL supply
To test whether the presence of non-AHL producing cells affects AHL mediated gene expression in the monitor strain, artificial microcolonies of the AHL monitor strain were constructed as above and overlaid with approximately 1000 live E. coli JM109 cells (Figs. 1 and 4). E. coli JM109 cells are not known to produce, degrade or respond to AHLs. The response of the underlying AHL monitor strain to 0.5 μM (non-saturing dose) and 5 μM (saturating dose) 3O-C6-HSL was then compared to control microcolonies lacking the additional cell layer.
In order to discount any optical dampening effects the additional cell layer might have had on image analysis of green fluorescence derived from the underlying cells, the control microcolonies were covered with a layer of E. coli cells immediately prior to imaging. Because this is a destructive process and the number of printed slides available for use in the study was limited, it was not feasible to compare time courses between controls and treatments. For this reason, GFP production from controls and treatments were compared at a single time point, when the GFP response to 3O-C6-HSL had reached a plateau (i.e., after 1 h). Comparisons between control and treatment slides were performed on four separate occasions. On each occasion, the average pixel intensity of 35 microcolonies from three fields of view was measured. Despite the fact that the absolute values obtained between replicates varied considerably, significant differences were observed between the fluorescence response of the treatment slides relative to the controls. The primary source of the absolute variability is likely to have been the exact cell density or physiological state of the cultures used to construct the microcolonies on different occasions.
As expected, the fluorescence response to 5 μM 3-oxo-C6-HSL was higher than for 0.5 μM 3-oxo-C6-HSL (Fig. 5). Interestingly, in the presence of E. coli JM109, the response of the AHL monitor strain to 0.5 μM 3O-C6-HSL was significantly lower than that of the control aggregates after 1 h (P= 0.039; Fig. 5). At a saturating concentration of 5 μM 3O-C6-HSL, however, the top layer of cells had no significant impact on the green fluorescence response of the underlying AHL responsive cells when compared to controls (P= 0.231; Fig. 5). These data suggest that non-AHL producing cells have an inhibitory effect on the AHL induced GFP response of the monitor strain at non-saturating concentrations, but no effect when the system is exposed to saturating concentrations of AHLs.
3.3Non-AHL producing cells increase the response of artificial microcolonies to an endogenous 3O-C8-HSL supply
In addition to testing the impact of non-AHL producing cells on the response of the monitor strain to 3O-C6-HSL supplied exogenously, we had the opportunity to determine whether or not the same effect would be observed if the source of the AHL was internal to the artificial microcolonies. To this end, homogeneous mixed species microcolonies were constructed consisting of an A. tumefaciens strain that overproduces 3O-C8-HSL, and the E. coli AHL monitor strain used above (Fig. 1). These microcolonies contained approximately 500 cells, half of which were AHL producing, and half of which were AHL responsive.
Microcolonies of this design generated green fluorescence without exogenous AHL addition. The fluorescence response was less intense than in the previous experimental design, presumably because there were approximately half the number of GFP producing cells within the microcolony. The response reached a plateau after two hours (data not shown).
Mixed species microcolonies were then overlaid with E. coli JM109 cells (Fig. 1). After a two hour incubation period, the impact of the covering layer of cells on fluorescence intensity was measured and compared with control microcolonies. As above, control and treatment slides were compared on four occasions, with the pixel intensity of 35 microcolonies in three fields of view being measured. Control microcolonies were prepared as for treatment microcolonies, but were not covered with E. coli cells until just prior to imaging as described above. In contrast to the previous experiment, the fluorescence response of the AHL monitor strain was significantly higher in treatment microcolonies when compared to controls (P= 0.025; Fig. 6). These data suggest that non-AHL producing cells can enhance the AHL mediated GFP response of the monitor strain induced by A. tumefaciens cells producing 3O-C8-HSL.
Traditionally, microbiologists have used a reductionist approach to great effect in the elucidation of molecular phenomena underlying microbial behaviour. Ongoing technological advances, however, provide opportunities to study microbial responses in the more ecologically relevant context of mixed species assemblages. In this investigation, we have employed dielectrophoresis to test the simple hypothesis that AHL induced phenotypes will be influenced by the presence of neighbouring non-AHL producing cells.
To this end, we constructed two distinct artificial microcolonies of AHL responsive cells and measured the AHL induced GFP response of these microcolonies in the presence and absence of non-AHL producing cells. In the first instance, when AHL was supplied exogenously, green fluorescence was reduced by the presence of non-AHL producing cells. In the second instance, when AHL was produced endogenously, the presence of non-AHL producing cells enhanced green fluorescence. We believe that these data falsify the null hypothesis that AHL mediated gene expression responses are necessarily identical in the presence or absence of non-AHL producing cells. The fact that green fluorescence was reduced in the first instance and enhanced in the second is a seemingly complex result worthy of discussion.
There are a number of reasonable hypotheses to explain how the presence of non-AHL producing cells reduced green fluorescence when AHL was supplied exogenously. The most compelling one of which would be that the covering layer of non-AHL producing cells creates a diffusion barrier that retards the movement of nutrients and/or oxygen into the AHL responsive microcolony, thereby interfering with the production or activity of the green fluorescent protein. Despite our best efforts to discount this phenomenon through the regular replenishment of media, it cannot be ruled out. A credible alternative hypothesis is that the layer of non-AHL producing cells acted as a barrier retarding the movement of AHLs into the AHL responsive microcolony, thereby reducing the AHL induced transcriptional response of the gene encoding the green fluorescent protein. The fact that a saturating concentration of AHL abolished the observed effect, suggests that the production or activity of GFP was not affected by nutrient or oxygen limitation, and therefore supports the latter interpretation.
Similarly, there are a number of hypotheses to explain how the covering layer of non-AHL producing cells enhanced green fluorescence in the AHL responsive microcolony when AHL was supplied endogenously by an A. tumefaciens strain. There is a guild of plausible explanations involving the A. tumefaciens strain reducing pH in local microenvironments  or producing a metabolite (besides AHL) or an enzyme that stimulates GFP production or activity. For example, a high affinity oxidase, such as the cytochrome cbb3-type oxidase produced by A. tumefaciens, could act to reduce the concentration of toxic reactive oxygen species within the microcolony, thereby enhancing the GFP response in the AHL monitor strain. Alternatively, it is again reasonable to hypothesise that the covering layer of cells acts as a barrier to AHL movement, thereby retarding the egress of the stimulatory metabolite from within the microcolony. The resulting endogenous accumulation of AHL would enhance transcription of the gene encoding GFP.
In summary, we have falsified the null hypothesis that AHL mediated responses are unaffected by the presence of non-AHL producing cells in close proximity. Whilst a number of explanations exist for the two results observed, we must, in deference to the principle of Occam's Razor, accept the single hypothesis that can explain both results. Namely, that non-AHL producing cells can act as a barrier to AHL movement. These results lend credence to the diffusion sensing hypothesis described by Schell  and Redfield , and raises the fascinating possibility that the selective advantage of AHL mediated gene expression mechanisms lies in the ability of signal concentration to reflect crowding by self as well as non-self cells.
It is important to note that this study does not address the ecological or environmental relevance of the observed phenomenon. It is unclear, for example, whether colonial architecture functionally similar to that achieved here by DEP exists, or is prevalent, in natural mixed species assemblages such as biofilms or flocs. Further, we have not tested other AHL producing or responsive strains to assess whether or not they would behave in the same manner as those used in the current study. Indeed, future research directions include validation of this model system by testing the impact on AHL mediated gene expression of crowding a range of different AHL producing strains with non-AHL producing strains, AHL degrading strains and inert materials such as beads and polymers.
We thank Jens Bo Andersen, Mike Givskov and Clay Fuqua for providing bacterial strains. We also thank Andrew Whiteley for conducting the flow cytometry analysis, Andy Lilley for aiding in image analysis and an anonymous reviewer who made excellent recommendations for improvement of the manuscript. This research was supported by studentships from the BBSRC and CASE support from the Centre for Ecology and Hydrology.