Heterogeneity in biofilms


*Corresponding author. Tel.: +49 (30) 314 73461; Fax: +49 (30) 314 73461, E-mail address: szewzyk@compuserve.com


Biofilms, accumulations of microorganisms at interfaces, have been described for every aqueous system supporting life. The structure of these microbial communities ranges from monolayers of scattered single cells to thick, mucous structures of macroscopic dimensions (microbial mats; algal-microbial associations; trickling filter biofilms). During recent years the structure of biofilms from many different environments has been documented and evaluated by use of a broad variety of microscopic, physico-chemical and molecular biological techniques, revealing a generally complex 3D structure. Parallel to these investigations more and more complex mathematical models and simulations were developed to explain the development, structures, and interactions of biofilms. The forces determining the spatial structure of biofilms, including microcolonies, extracellular polymeric substances (EPS), and channels, are still the subject of controversy. To achieve conclusive explanations for the structures observed in biofilms the cooperation of both fields of investigation, modelling and experimental research, is necessary. The expanding field of molecular techniques not only allows more and more detailed documentation of the spatial distribution of species, but also of functional activities of single cells in their biofilm environment. These new methods will certainly reveal new insights in the mechanisms involved in the developmental processes involved in the formation and behavior of biofilms.


Surface-associated microbial communities have been investigated since newly developed optical instruments permitted a direct view into the microcosm. Starting with Leeuwenhoek and his description of the small ‘animalcules’, generations of microscopists have observed and described the organisms in free water and also on surfaces. An early technique for the examination of surface-associated microorganisms, was to expose glass slides in all types of aquatic and terrestrial habitats [1,2]. The developing communities, named ‘Aufwuchs’ by German microscopists, have been examined scientifically many times up to the last third of the 20th century, and are still interesting subjects for amateur microscopists. These early examinations of Aufwuchs focused mostly on the sessile protozoa and lower metazoans as well as on the algae. Bacteria were mentioned in only few studies until the beginning of the last century when several microbiologists started systematic examinations of the occurrence of bacteria on surfaces and of the effect of surfaces on bacterial physiology [3,4].

Even in drinking-water distribution systems, basically all existing interfaces are covered by biofilms, from the waterworks through the distribution system to all house installations. The existence of these biofilms was reported and documented several decades ago [5]. These descriptions of drinking-water biofilms were mostly based on microscopic and/or electron microscopic examinations of pipe materials and encrustations from pipe surfaces.

In the 1970s and 1980s many attempts were made to develop mathematical models to describe processes in biofilms. To make the equations more tractable, several simplifying assumptions were made. One of these was that biofilms were assumed to be homogeneous structures. With this simplification, the calculation of diffusion gradients was possible. As one consequence, experimental setups were developed which allowed the growth of biofilms of defined thickness and of a relatively homogeneous structure (e.g. Rototorq, Constant Depth Film Fermenter) or synthetic biofilms were constructed by immobilization of bacterial cells in gels.

During the last two decades, biofilms have been recognized as highly structured habitats to be found at almost every phase interface, as long as it is exposed to water for at least some time. In marine and freshwater environments, biofilms were discovered to harbor unusual, sessile bacteria, as well as stages of the life cycles of bacteria so far considered to be planktonic. The advantage for an organism of being attached to a surface was first discovered in oligotrophic systems, where the ‘antenna’ effect of the surface provides a more suitable environment for bacterial growth [3,6–8].

It was also shown that the close spatial arrangement of different bacterial species can be advantageous to the community as a whole, for example in the degradation of recalcitrant organic molecules. For the complete degradation of such substances, the interaction of several species (co-metabolism) is the most efficient way and functions best if the species involved are spatially close together. The advantage of such associations and aggregates for degradation accounts for their wide use in all types of wastewater treatment (activated sludge, fixed-film reactors, and up-flow anaerobic sludge blanket (UASB) reactor).

With the development of new tools for the in situ examination of biofilms the structural heterogeneity of natural biofilms, already described in principle by the ‘old microscopists’, was demonstrated in more detail, including the existence of channels in the film and of a broad variety of surface structures. Furthermore, the application of microelectrodes with resolutions of a few microns and of genetic reporter systems revealed that the spatial heterogeneity is accompanied by a physiological heterogeneity.

2Biofilm structure

Throughout the history of ‘biofilm’ it ought to be clear that the term embraces a huge collection of different structures. Any microbial organization which can range from patchy monolayers on some surfaces through very thick gelatinous masses associated with water cooling systems to filamentous accretions near sewage outlets probably needs a more sophisticated definition than the panchestron which is biofilm! Science has historically tried to make simplifying assumptions concerning systems which are ill understood. The simplifying assumption that probably extended through the 1980's well into the following decade was that a biofilm could be represented as a simple planar structure, largely 2D, with a relatively constant thickness. Examples of such systems were presented graphically by Costerton et al. [9]. It is worth stressing here that other patterns were clearly recognized by the end of the 1980s. For example Wilderer and Characklis ([10]) showed pictures of filamentous as well as substantially planar biofilms in the proceedings of a Dahlem Conference.

New technologies, including the confocal laser scanning microscopy (CLSM), the differential interference contrast (DIC) microscope, microelectrode and a number of other techniques, threw serious doubt on the picture of a largely homogeneous flat structure.

At length the planar biofilm consensus gave way to a number of new views, and it became clear that there were at least three different structural ‘camps’. The first was the traditional, planar, homogeneous view of a biofilm structure. This view came from dental researchers, largely from transmission electron microscopy (TEM) of plaque samples. Examples of such biofilms were published by Nyvad and Fejerskov [11].

A second viewpoint was that of Keevil and his colleagues [12] using DIC microscopy to examine samples growing on the inner surfaces of water distribution systems. These workers discerned stacks consisting of microcolonies of bacteria held together by extracellular polymeric substances (EPS) and appearing as columns surrounded by a liquid phase in which grazing protozoa could be discerned. Below the stacks there appeared to be a layer of cells about 5 μm thick attached to the substratum. These types of structure led Keevil to name this the ‘heterogeneous mosaic model’.

The third type of biofilm has led to what is perhaps the current icon of microbial biofilm, the mushroom or tulip model. Here CLSM coupled with the use of fluorescent markers has revealed an interesting common structure, that of the biofilm forming as mushroom shaped objects with stalks narrower than the upper surface parts, the whole penetrated by channels through which a liquid phase is free to move with the prevailing flow. Such pores and channels have been seen in many different types of biofilm. The importance of these channels has been stressed by different workers. That they carry oxygen around and between the mushroom stacks is not in doubt thanks to work with oxygen microelectrodes by de Beer and colleagues [13]. Flow through the channels could also be monitored and flow rates determined using time lapse photography to follow the movement of fluorescent beads [14]. Much of this work was carried out using a laboratory model system fed with acetate or other carbon sources at low concentrations.

There is one unifying thread that can be discerned in these three types of structure, and that is substrate concentration.

The heterogeneous mosaic model was formulated on the basis of growth systems fed with unamended water as party of a research project on biofilm growth in water distribution systems. The concentration of organic substrates in such waters is very low and probably measured in units of μg or mg l−1. In the oral environment nutrient concentrations vary throughout the normal day. At their highest carbohydrates can reach 100 g l−1 during consumption of very sweet foods or confection. During the remainder of the time nutrient concentration and type alter. The oral flora relies for much of the time on glycoproteins, especially mucin, which are produced by the salivary glands in the mouth. The picture is a little more difficult to interpret because of these generally regular perturbations.

Mushroom structures are formed at intermediate substrate concentrations, generally in the laboratory where nutrients are added as a matter of course.

A simple interpretation is that substrate concentration must play some part in determining biofilm structure. The first observations on the regulatory effect of substrate concentrations on the dynamics of biofilm structure were reported from studies with anaerobic bacteria forming colonies in continuous flow cells with varying substrate concentrations [15]. In this study it was demonstrated that colony size directly correlated with the substrate concentration in the flowing medium. The substrate concentration dependent mechanisms regulating colony size were growth and detachment of motile swarmer cells. In other studies biofilm structure was related to both substrate concentration and hydrodynamic conditions [16] and the view expressed that biofilm structure was determined by resource concentration as well as by the shear forces applied to the community. This work was made slightly harder to interpret since the authors considered shear force and substrate concentration as two dependent factors which could be represented on a 1D array, whereas it might be preferable to consider these two as entirely separate factors, which they obviously are.

3Structure as revealed by microscopy

There are a growing number of microscopy techniques that have provided a massive amount of information on biofilm formation. All have some problems in the interpretation of the images they produce, since the procedures used are capable of generating artefacts which are then revealed in the final pictures. For example the EPS which generally surround and engulf microbial communities can dry down to fine strands which can be interpreted as fibrous structures attached to the organisms themselves.

3.1Traditional light microscopy

The complex structure of biofilms has already been revealed by light microscopy [1]. Based on morphological characteristics, a large variety of unusual bacteria was described, indicating a high species diversity [17].

3.2Electron microscopy


Perhaps the most traditional approach to electron microscopy relies on fixing, staining and sectioning material. This has been used by many workers. Some good examples come from the examination of oral microbial communities [11]. Using this technique allows an examination of cells: their morphology, disposition and something of their ultrastructure. In addition use of specific dyes can reveal EPS, in particular polysaccharides using ruthenium red, whilst gold-labelled antibodies can be powerful tools in identifying specific bacteria. There may be artifacts here too; EPS is certainly not in its natural form due to dehydration, and there is some doubt about the identity of ‘mesosomes’ found within certain cells. In general, with a few caveats, TEM has been and still is a powerful tool in the examination of biofilm structure.

3.2.2Scanning electron microscopy (SEM)

SEM can also give useful information on the surface structure of a biofilm. After fixing and shadowing it is almost impossible to see the ‘natural’ surface, and this is in some ways a good thing. What can be seen are the cells and their disposition. If EPS existed, it dries quickly leaving visible only thin threads, and leading to some structural misconceptions as mentioned earlier. If the microscope is equipped with low vacuum accessories, so that fresh samples can be examined without further treatment, the surface of most biofilm resembles cloud cover. This is the biofilm and its associated EPS. The result is conceptually informative if nothing else.

3.3Atomic force microscopy (AFM)

AFM is another powerful tool which in some applications can reveal atomic structure. On the biological scale it can also reveal the surface of a biofilm, once more showing a relatively smooth undifferentiated surface due to EPS formation.


As is often the case in science, a very simple idea revolutionized microscopy and provided perhaps the single most powerful tool in experimental biofilm research. The concept behind CLSM was that shining a coherent light beam onto a specimen and detecting the in-focus light from it via a pinhole would generate a clear point image since all the out of focus information would be eliminated by the pinhole. If the beam scans across the specimen then a 2D image is formed. The geometry of the system can be altered so that the focal plane can be at different depths in the specimen, at least as far as the light beam can penetrate. Using image processing techniques all the 2D images can be stacked and a 3D picture be produced for the specimen. Though a powerful tool in its own right, the development of very specific fluorescent probes added a high degree of specificity to CLSM allowing the operator to detect and identify individual organisms from phylum right down to strain, depending on the types of probe used. In addition, other probes could be used to determine aspects of the biochemistry of the organism and the physico-chemistry of its immediate environment.

One of the earliest probes to be used was fluorescein. This provided a fluorescent background against which bacteria could be viewed. There are probes which can be used to distinguish between Gram positive and Gram negative bacteria. Resazurin has been used as an indicator of biochemical activity. This has been used as an indicator of ‘life’ or viability. Of course such a definition is fraught with difficulties, since a cell incapable of reproduction might still have the ability to respire. Other fluorescent cocktails are marketed to differentiate between live and dead organisms though the structures mentioned above still broadly apply. Baclite ‘Live Dead’ is based on membrane integrity, a useful indication of viability, though not infallible either.

The attachment of fluorescent moieties to other components can increase their versatility. Linked to dextrans of different molecular masses they can help determine diffusion rates as well as charge distribution using polyanionic dextrans. Conjugated to lectins they can be used to determine the distribution of oligosaccharides. Several approaches allow fluorescent molecules to be used to detect different species. Fluorescent dyes conjugated to specific antibodies provide a powerful technique, although not so valuable as the use of the highly specific 16S rRNA probes. 16S rRNA is present in all living cells and is almost ideal for the job because of its molecular size and the mixture of highly conserved and variable regions in its structure. Probes to the more conserved regions can be used to determine whether an organism is from the Bacteria or a member of the Archaea. Other probes can be increasingly specific, recognizing families, genera, species and even strains of the same species. Fluorescent probes can respond to environmental composition. Thus, carboxyfluorescein has been used to determine local pH in Vibrio parahaemolyticus biofilm communities [18].

A recent advance in CLSM technology is promising exciting results in biofilm research. This is the multiphoton confocal microscope. In this instrument fluorescence is induced by the simultaneous action of two or more photons generated by a powerful infrared laser operating in pulse mode at ns intervals. The infrared beam has a number of advantages: (i) it penetrates further into the sample, up to 10 times the distance of the normal CLSM laser; (ii) because of the longer wavelength photo-bleaching is much less serious a problem; (iii) pulsing the beam allows the determination of fluorescence decay rates independent of fluorophore concentration. This last effect has been used by Vroom et al., [19] using an oral community grown in the Constant Depth Film fermenter. pH values around certain cell clusters were as low as 3.0 up to 140 μm into the biofilm. CLSM has been used to measure liquid flow rates in heterogeneous biofilm. In this case fluorescently labelled latex beads were tracked at intervals as they moved through the pores and channels of a biofilm [14].

So what has microscopy told us about biofilm heterogeneity? Traditional light microscopy has revealed a family of different types of biofilm, although an early ‘consensus’ model was that biofilm could be regarded as a substantially flat, relatively homogeneous material. Examination of water distribution systems using light microscopy, in particular DIC microscopy, showed that microcolonies were often well separated ‘stacks’ around and through which protozoa could be seen. CLSM has given us the clearest picture of biofilm structure and led to a new consensus that many biofilms are based on tulip- or mushroom-shaped structures whose upper parts can often coalesce leaving ‘galleries’ or channels below through which environmental fluids can move, often acting as transport vessels delivering nutrients, removing waste products and acting as conduits for messenger molecules.

4Structure as revealed using microelectrodes

Biofilm communities range in depth from bacterial monolayers, perhaps 1–2 μm thick up to a few hundred μm, and often more than this. Structures within biofilms are quite small in scale so that, if it is necessary to determine some of the physico-chemical variables on this scale, appropriate sensors are needed. Fortunately there is an increasing range of microelectrodes that are appropriate to the problem. Microelectrodes have been used by microbiologists for around 20 years. For example, Chen and Bungay [20] examined oxygen penetration into trickling filter ‘slimes’ and Wimpenny and Coombs [21] used oxygen microelectrodes to determine the depth to which oxygen penetrated into bacterial colonies. Revsbech [22] has reviewed the application of microprobes to biofilm. The irregular structure of most biofilms revealed by CLSM suggested that substrate transfer was not due to molecular diffusion alone but included a convective term as well. It seemed sensible to examine the levels of oxygen at different points within a biofilm. DeBeer et al. [23] performed this experiment and reported that if the tip of an oxygen microelectrode was inserted into a colonial aggregate within the biofilm, the PO2 dropped to zero. However deeper within the film, within a water channel there was measurable oxygen present. The conclusion could only be that there was a convective liquid path through the structure.

A range of different microelectrodes have been developed capable of measuring nitrous oxide and oxygen [24]; oxygen, H2S and pH [25]; nitrification substrates and products [23]; sulfate reduction and sulfide oxidation [26].

Use of microelectrodes plus microslicing and other chemical and physical techniques [27] showed that a wastewater biofilm is highly stratified. As depth increases so, too, does density whilst porosity, metabolically active biomass and oxygen diffusivity all fall.

5Community structure as suggested by molecular methods

The problem that about 99.9% of the bacterial cells in biofilms from every ecosystem studied could not be cultured on standard media remained unsolved until new approaches were made to examine the population structure, for example, in drinking-water systems.

The application of molecular tools, especially in situ hybridization with 16S or 23S rRNA-directed oligonucleotide probes, was a starting point to resolve this problem. In situ hybridization allows the detection of specific nucleic acid sequences in eukaryotic and prokaryotic cells by binding of oligonucleotide probes to their complementary target sequences. Comparative sequence analysis of small subunit rRNA provided not only the basis for a bacterial taxonomy based on an assumed natural relationships of the prokaryotes [28,29], but also the development of synthetic oligonucleotide probes which are specific for defined phylogenetic groups of organisms [30]. Fluorescence in situ hybridization (FISH) in combination with advanced microscopic techniques such as CLSM, deconvolution microscopy and digital image analysis, became an important part of a polyphasic approach in microbial ecology for the in situ identification and localization of microorganisms within complex environments [31].

A successful example of this approach was applied to young biofilms from the drinking-water supply in Berlin. Phylogenetic staining of the biofilms with FISH probes directed against the major sub-classes of the Proteobacteria [32] revealed that about 80% of the biofilm bacteria present in the in situ experiments were β-Proteobacteria [33]. This led to a strategy, which focused on colonies formed by members of the β-subclass of Proteobacteria and isolation of only these strains [34]. Oligonucleotide probes to different classes of bacteria and to each of the main isolates were then obtained or prepared. The probes were used to check all the isolates and deployed to biofilm in situ to determine what proportion of the natural community could be isolated. Using the eight probes constructed from the original isolates, from 66–88% of the natural population belonged to classes represented by the probes [35,36]. This was one of the first examples of an accurate assessment of nearly all the members of a natural microbial population.

6Structure as suggested by mathematical modelling

There are about as many different approaches to investigating biofilm structure as there are scientists studying them. Apart from the experimental methods discussed earlier, there is a whole array of methods which involve the use of mathematical and computer-based models. For a guide to some of the types of mathematical model that can be applied see Characklis [37].

The mathematical route can be broadly divided into two major branches, ‘continuum’ and ‘discrete’ models. Continuum systems reflect a ‘traditional’ mathematical methodology relying on the solution of families of differential and partial differential equations. Such an approach is associated more with large scale processes where there is some sort of homogeneity. Continuum models are used to model mineral cycles in the oceans, growth dynamics of microbial communities, and in some cases biofilm activities themselves. Often they are applied by chemical and process engineers to predictions of the metabolic rate constants of economically important biofilm, for example associated with the water industry in effluent treatment. Two examples of continuum models are instructive: one from the process engineers, the second one from oral microbiology.

Most early research, certainly until the late 1980s, assumed that microbial films were smooth planar homogeneous structures and this led to the formulation of continuum mathematical models such as ‘Biosim’ (Wanner, 1989). Much of this work was related to a need expressed by chemical and water engineers for predictions about the rate of biofilm chemical activities, as well as a need to determine physical constants associated with biofilm thickness, such as heat conductivity, friction coefficients and the effect of biofilm on flow and pressure changes in water conduits. The realization that many biofilms were far more irregular in structure, led to the development of a more sophisticated version of ‘Biosim’ called ‘Aquasim’[38]. The latter allowed for the attachment and detachment of cells at the substratum surface, and at the same time could tackle some of the problems of spatial heterogeneity, such as transport through pores and water channels.

Another example of the application of continuum models to biofilm research is due to George Dibdin and his colleagues who were mainly concerned with pH changes which play a dominant part in the biology and clinical importance of dental plaque. Shellis and Dibdin [39] noted that such oral communities had a very high buffering capacity. This led Dibdin [40,41] to formulate a model for ion movement in dental plaque. This incorporated fixed charges associated with polymeric materials in the plaque, as well as neutralization of acidic groups by carbonate and phosphate which diffuse into the plaque from saliva. This system has recently been reviewed [42].

Discrete models operate in quite a different manner. These become necessary where the scale is such that individual events are important, rather than assuming some kind of global, larger-scale, average activity for which continuum models are satisfactory. Very few protons determine the pH around a microbe, a number of microbes interacting, the behavior of individuals from a group, all these are fit subjects for discrete modelling systems.

Cellular automata are one form of discrete modelling. They consist of an addressable array on which objects are located. In microbiological applications the latter can include microbes and solute molecules, for example substrates and products. A cellular automaton (CA) operates using rules which can be quite simple. These act ‘globally’ over the whole array but are applied ‘individually’ in each compartment. The process is iterative, the computer examines and acts on the content of each compartment in turn, as often as is necessary. The Game of Life was one of the earliest CA models. Here there are only a few simple rules: cells live, divide or die according to the occupancy of spaces adjacent to them. CA models have so-called ‘emergent’ properties. That is they can generate complex patterns that cannot be predicted any faster than by running the simulation itself.

A simple CA model was developed by Wimpenny and Colasanti [43] to model microbial growth in biofilms. An array was first populated with ‘substrate’ molecules. These were allowed to move randomly around their original compartment at each iteration of the program. This is entirely analogous to molecular diffusion though in the model diffusion was on a much larger scale. ‘Cells’ were ‘inoculated’ onto the bottom edge of the array (the ‘substratum’). Each cell required substrate to grow and divide providing there was a free space adjacent to it. Array size, substrate diffusion rate and growth yield could all be varied.

The CA model was able to replicate some of the experimental results albeit rather crudely. At very low substrate levels independent stacks of microbes formed. These were extremely substrate-limited, as could easily be seen by the distribution of substrate units in the array. As solute concentration was raised so the growth pattern became denser and in some simulations at appropriate diffusion rates and yield coefficients mushroom structures and water channels could be seen. At the highest substrate levels used in the simulation a dense uniform biofilm appeared. These simulations clearly demonstrated that one explanation for the structures observed was that the growing biofilm depleted a zone around it so that at low concentrations there was a ‘no go’ zone around the community. As the concentration and flux of substrate molecules increased there was a change from substrate limitation to growth rate limitation and the cells could occupy more and more of the available space until at some point confluent growth was observed.

These modelling results were confirmed by Picioreanu et al. [44] using a much more realistic model, and later by Hermanowicz [45] also using a CA. Picioreanu's model system though built around a CA, was in fact a hybrid model. Continuum equations were used to calculate solute distribution and growth rate. More recently this group has developed a more sophisticated hybrid computer model which uses a range of numerical methods to reproduce solute diffusion, shear and detachment forces, biomass growth and cell distribution which is used to calculate the distribution of biomass.

Another related approach to discrete modelling is the use of autonomous agents. These are sometimes known as ‘intelligent’ agents and are examples of object-orientated programming. Each agent represents a piece of computer code. Perhaps the best developed of this type of programming is the Swarm system generated by computer scientists at the Santa Fe Institute in New Mexico, USA. Here agents may be modelled collectively as ‘swarms’, literally like a swarm of bees. The concept is iterative since a swarm can itself be an agent of another larger scale swarm. This system is obviously well suited to collections of items. For example it has been used to model ant behavior and has significant applications in the ecological sciences.

Kreft et al. [46] have used the swarm system to model the growth of a bacterial colony. Here each bacterium (based on kinetic constants applying to Escherichia coli) is one agent. The cell has a very simple ‘physiology’. It transports substrate, metabolizes it, excretes product, has maintenance energy, grows or dies depending on nutrient availability and at a critical size divides. It then ‘shoves’ other cells apart to generate space for itself. If it dies it is recycled to provide extra resource.

More recently the same system has been applied to biofilm (Kreft et al., in press). The system, still under development, is based on a nitrifying community of two organisms using oxygen and ammonia and generating nitrite and nitrate as a final product. In addition the program can model the production of EPS under different alternative conditions (i) as a capsule to the producing cell; (ii) generally liberated into the environment where it pushes cells apart; (iii) generally liberated where the cells have some mechanism to stick together.

The swarm suite used by Kreft is a hybrid system as was that of Picioreanu. In both cases numerical equations were used to model solute diffusion. Our own feeling is that such modelling programs have a considerable amount to offer in microbiological research since they are capable of incorporating different community members each with its own characteristics. From rather simple rules complex structures with many similarities to natural communities, can emerge.

This has only been a brief description of the type of discrete modelling that can be applied to biofilm structure. Results, however, are unanimous in illustrating the effects of nutrient concentration as well as hydrodynamic processes on biofilm structure and any global picture of biofilm structure ought to take these processes into account.

7Metabolic and activity heterogeneity

The rapid development and application of new molecular tools in combination with advanced microscopic techniques not only provides detailed 3D images of the structure of biofilms, but also of the activity and physiological status of individual cells in the biofilm.

In addition to the inherent phylogenetic information of the FISH technique, the correlation between growth rates of bacterial cells, their average ribosome content, and the strength of the probe-conferred fluorescence can be used to determine the metabolic potential of individual bacterial cells [47]. This approach was applied in studies dealing with the activity of microbial cells in drinking-water biofilms. The detectability of cells with FISH in the biofilms was high (up to 80%) only in young biofilms (<3–4 weeks), while in old biofilms (>5–7 weeks) less than 50% of the DAPI-stained cells were detectable with FISH. To test for the in situ potential of individual cells to metabolize nutrients, the FISH-technique was used to evaluate the physiological status and the in situ growth potential upon addition of additional substrates [33]. Interestingly, members of one ‘species’ expressed a broad variety of activity patterns (measured as increase of ribosome content, cell division), although the cells were spread over the surface in a loose monolayer community with no obvious differences in the environmental conditions for the different cells.

A further new possibility in studying biofilm communities is the integration of reporter genes under the control of defined or undefined promotors. The ‘classical’ reporter genes like lux and gal have been used in many different systems. A newer reporter system, the green fluorescent protein (GFP) present naturally in the jellyfish Aequorea victoria can be used as a marker in bacteria. Here the necessary complement of genes can be transferred to the recipient cell either so that GFP is expressed constitutively, or under the control of a promoter associated with a particular function.

To evaluate the biofilm-dependent regulation of gene expression in E. coli K-12, random insertion mutagenesis with MudX, a mini-Mu carrying the promoterless lacZ gene, was performed and the gene expression in planktonic versus attached cells was monitored. Major changes in the expression pattern were observed during biofilm formation, and the transcription of 38% of the genes was affected [48].

The use of phylogenetic probes together with reporter genes can yield immensely useful results, especially when information on activity and/or physiological interactions becomes available.

Nielsen et al. [49] used GFP and a fluorescently labelled ribosomal RNA probe to investigate interactions between a Pseudomonas sp. and a Burkholderia sp., which could be differentiated by staining. The two were grown together in a slide culture. On citrate as a substrate, which could be utilized by both strains, the two grew forming separate microcolonies. Behavior was different when both strains were supplied with chlorobiphenyl as substrate which is only mineralized by both strains together in a metabolic loop with chlorobenzoate as intermediate. The net result was mixed colonial community! The conclusion must be that biofilm heterogeneity is dependent on many factors, including the nature of the resources and possible interactions leading to their use.

In another elegant study, the distribution of Diclofop degrading bacteria in a complex biofilm community was described. Catabolic enzymes for the degradation of related compounds were detected by PCR and bacterial isolates harboring these genes were subsequently detected in the biofilm community with specific oligonucleotide probes for FISH [50].

8Cell–cell signalling and biofilm structure

Cell–cell signalling has been recognized as an important feature of microbial communities and recent results show that it might be an important factor also in biofilm regulation.

Two quorum sensing systems have been reported for Pseudomonas aeruginosa. These are lasR-lasI which controls virulence but also regulates the expression of a second system, rhlR-rhlI which is involved in the formation of some secondary metabolites. Production of the signal molecules are controlled by these systems, butyryl homoserine lactone by rhlI and 3-oxododecanoyl-homoserine lactone by lasI. The wild-type and a mutant both attach to surfaces, however the lasI mutant can form only microcolonies and cannot progress to forming a mature, differentiated biofilm, unless the signal molecule 3-oxododecanoyl-homoserine lactone is added to the system. Obviously, signal molecules can be involved in some aspects of the control of biofilm formation [51].

Also using P. aeruginosa, O'Toole et al. [52] reported that crc mutants formed only a simple monolayer film instead of the denser structure produced by the wild-type. Since crc is associated with catabolite repression, the latter is clearly also implicated in biofilm formation. Loo et al. [53] isolated 18 mutants of Streptococcus gordonii which could not produce a biofilm. Nine of these mutations were linked to quorum sensing, osmoadaptation and signal transduction whilst the remainder had no known function, implying that there were several other unknown control systems associated with the formation of biofilm.

9Biofilm structure – a consensus

From all the experimental and theoretical evidence adduced above, one can derive some sort of consensus view as to the structure of biofilm. Many different factors are important. Some of these are listed in Table 1. It may not always be easy to put these factors in order of importance, and the search for a consensus on what regulates biofilm structure might be fruitless. Nevertheless there are a few indications.

Table 1.  Summary of the factors influencing the formation of biofilms at different times
Genotypic factorsThe specific genotype of the organism.
 Expression of genes encoding surface properties.
 Expression of signalling systems.
 Formation of EPS.
 Organism growth dynamics; specific growth rate, lag periods, affinity for substrates, yield coefficients etc.
 Expression of genetic factors not directly connected to biofilm formation (motility and chemotaxis, catabolite repression genes etc.)
Physico-chemical factorsPhase interface (combinations of solid, liquid and gaseous).
 Substratum composition and roughness.
 Substrate composition.
 Substrate concentration/gradient.
 Temperature, pH, water potential, pressure, oxygen supply and demand, radiation effects.
Stochastic processesInitial colonization: attachment, detachment.
 Random changes in biotic and abiotic factors.
Deterministic phenomenaSpecific interactions between organisms: competition, neutralism, cooperation and predation.
Mechanical processesShear due to laminar or turbulent flow conditions; abrasion; logistic restrictions.
Import–exportAddition or removal of biotic or abiotic components to a biofilm system, E.g. the import of sand, clay minerals or organic detritus into a biofilm structure. Sloughing off of biomass, release of individual (swarmer?) cells.
Temporal changesDiurnal or annual periodic changes in biotic and abiotic environment, e.g. light, temperature, pH, PO2. Irregular changes due to unforeseen events.

The initial colonization of a clean surface, although restricted to a subset of the whole population that can colonize surfaces, will be entirely random, depending on what lands where and when. Any juxtapositions which might have subsequent consequences are initially quite fortuitous. Once organisms have ‘landed’ any possible interaction will happen since at this point they are governed not by chance but by a set of physico-chemical and biological rules. Two microcolonies might compete, cooperate, or be independent.

Clearly, the developmental potential of the organism, as determined by its genome, is one of the most important determinants of structure. Thus, filamentous organisms might be expected to form a different multicellular array than would nonmotile coccoid organisms. Perhaps the majority of biofilms can be regarded as approximately planar systems which include microcolonies and more or less abundant EPS, and this would differentiate them from clearly filamentous structures which might include cell masses formed by, e.g. Sphaerotilus natans. Such a structure we now know will be variously penetrated by channels and pores.

At this point there are two competing views regarding the spatial organization of the biofilm. The first is purely physico-chemical and is based on diffusion-limited aggregation models. This is based on the common sense precept that as substrate concentration is lowered any microbial accumulation will remove it from its own neighborhood, rendering an adjacent area unsuitable for further growth by competing cells or aggregates of cells. The evidence for this is partly a comparison of different experimental observations and partly through the unanimous view of all those who have applied discrete modelling systems to the problem. Another view, discussed earlier, is that such morphology is under the control of the genome of organisms involved, and there seems to be experimental evidence that this is true. Mutants with defects in appropriate signal systems seem to have lost the ability to produce the sort of channeled array that is the ‘badge’ of current thinking on biofilm structure.

As a matter of fact, we see no serious problems with these two views. If the fundamental process of substrate diffusion limitation provides a skeleton on which biofilm formation takes place, there is nothing to say that evolution and natural selection should not have taken advantage of this where appropriate in forming a structure with some additional benefits. For example, a channeled structure has an obvious advantage in that it allows fluids to pass through, supplying substrates and removing products which otherwise would have had to move by molecular diffusion alone, a much slower process.

Perhaps next in the list of environmental determinants must be mechanical forces, in particular shear which will occur in many natural environments where liquids, generally water, are flowing across a substratum surface. Shear forces remove sections of biofilm, but also have a profound effect on the overall structure leading to a flatter and smoother biofilm. Projections from such a structure form streamers which can oscillate in the flow and are finally removed. This process is combined with active detachment of cells from the biofilm. In a natural ecosystem the biofilm will be exposed to organic and inorganic detritus some of which will become incorporated into the structure. In addition, everything said so far is governed by a variety of physico-chemical factors such as pH, temperature, osmolarity, radiation, light, oxygen tension as well as by possible regular or irregular shifts of these factors.

10Community dynamics and evolution – one possible origin of metazoans?

That a biofilm may often be a community with emergent properties, raises intriguing theories concerning the origin and evolution of metazoans.

The first paleontological evidence for primitive life are remnants of microbial cells in layered structures, fossilized biofilms. These biofilms probably represent microbial mats or stromatolites, complex microbial associations already requiring different types of inter- and intraspecies interaction. The first preserved remnants of metazoans are fossilized sponges. Paleontologists and zoologists, studying the development and evolution of sponges, noted the morphological similarities between some free-living flagellates, choanoflagellates, and specialized cells in sponges, the choanozytes. This observation led to various hypotheses, assuming that sponges might have evolved from multicellular colonies of choanoflagellates.

Additionally, the histological examinations of an increasing number of sponge species resulted in the discovery that many sponges harbor populations of microorganisms, especially bacteria, in their tissues. The density and relative proportion of bacteria in the total sponge biomass differed between the various species. In some species only a few microorganisms were detectable, while in others up to 60% of the total sponge biomass was made up by associated bacteria. The hypotheses on the origin of sponges were therefore modified, assuming that the free choanoflagellates were living in and together with microbial biofilms. During the further evolution the association between bacteria and flagellates, both being part and constituents of the biofilms, might have further increased in complexity, finally resulting in the institutionalization of the microbe–flagellate-association as primitive types of sponges. Possible links between the stage of colonies of choanoflagellates (e.g. Proterospongia) and primitive sponges are known. However, the occurrence of bacteria in such colonies or organisms has been studied only very superficially since attention was paid primarily towards the eukaryotic cells, and bacteria were often only seen as accidental structures, if noticed at all.

The possible involvement of bacterial biofilms in the early evolution of metazoans is a tempting idea, assuming that heterogeneity, complexity and interaction in microbial biofilms lead to the development of sponge structures and subsequently of the existing diversity in metazoan life.

A further intriguing observation is the finding of increasing numbers of new secondary metabolites and lead structures in marine invertebrates, especially sponges [54,55]. It was therefore hypothesized that the production of the secondary metabolites might be a consequence of the coexistence of eukaryotic and prokaryotic cells in a highly ordered association.

11Biofilm heterogeneity: the future

During the last decade we have learned a lot about species diversity and the spatial distribution of different species in biofilms. The challenge for the future is to understand the patterns in gene expression, underlying physiological properties and the interaction between cells in biofilms.

The fast development of molecular tools has opened a new field to study in detail the physiological activity and status of individual cells in a spatial order. As a consequence, the regulatory mechanisms (e.g. sensing, signalling) will be further explored to understand the physiological and morphological potential of a ‘species’. These studies will not only illuminate the complexity of biofilm communities but will also help to understand basic questions like: ‘When is a cell dead?’, ‘What does the viable-but-non-culturable state mean in the life cycle of bacteria?’. The potential of bacteria to undergo complex morphological differentiation, always accompanied by physiological adaptations, should be addressed in regard to the close proximity of biofilm communities. In addition, the role of purely physico-chemical factors in regulating biofilm structure needs to be considered in tandem with genotypic factors. Finally these studies, in combination with more and more sophisticated mathematical models, may even lead to a unifying theory of not only cell but also biofilm developmental cycles.