Fluorescence-labeled oligonucleotide probes were applied, combined with in situ reduction of the fluorochrome 5-cyano-2,3-ditolyl tetrazolium chloride (CTC), to describe the development of bacterial density, phylogenetic diversity and bacterial metabolic activity during the formation of drinking water biofilms. Polyethylene and glass surfaces exposed to drinking water in a modified Robbins device were rapidly colonized by a biofilm community of phylogenetically diverse prokaryotes, and cell density of the biofilm community was strictly controlled by grazing eukaryotic organisms. In situ hybridization with group-specific rRNA-targeted oligonucleotide probes revealed the following: (i) the prevalence of bacteria belonging to the β-subclass of Proteobacteria within the bacterial biofilm populations; (ii) differences in the population composition, assessed by phylogenetic probes, depended on the surface properties of the substrata; (iii) the influence of water retention time on variations in population structure; and (iv) the presence of bacteria belonging to the family Legionellaceae associated with grazing protozoa. The metabolic potential of bacteria was assessed during biofilm formation using fluorescence signals after in situ hybridization and the reduction of the redox dye CTC as an indicator of respiratory activity. Respiratory activity and ribosome content of adherent bacterial cells decreased continuously during the early stages of the biofilm. After 35 days the percentage of CTC-reducing cells stabilized at 30%, and the amount of hybridized cells stabilized at 55%, of the initial cell number. To ascertain the amount of dormant, but potentially active cells, we established a new method, defined as probe active counts (PAC). Biofilms were incubated with a mixture of appropriate carbon sources and an antibiotic preventing bacterial cell division, followed by the determination of metabolic activity by in situ hybridization. By this approach the percentage of hybridized cells could be increased from 50% to 80% of total bacterial cell counts in the oligotrophic drinking water biofilms.
Biofilms are common in all drinking water distribution systems and in the past, different aspects of biofilm formation have been examined [1, 2]. The general structure of biofilms has been studied using light or scanning electron microscopy [3–5] and several biofilm forming bacteria have been characterized and classified using conventional culture techniques [3, 6]. However, traditional culturing approaches are biased by the intrinsic limitation towards easily cultivable and fast growing bacteria and might not reflect the true bacterial composition of natural biofilm communities.
Thus, most surveys in drinking water systems have not attempted to describe the whole spectrum of organisms, but have focused either on species causing infectious diseases such as Legionella pneumophila, opportunistic pathogens like Mycobacteria spp.  or indicator organisms for fecal contamination, such as coliform bacteria .
Nowadays, rRNA-targeted oligonucleotide probes are established as in situ tools in microbial ecology and are increasingly used to identify bacteria in their natural habitats [10–14]. Recently, we demonstrated the applicability of fluorescent oligonucleotide probes in oligotrophic environments, such as drinking water system biofilms [15, 16]. Beside providing phylogenetic information, oligonucleotide probes can be used to assess the metabolic status of single bacterial cells, based on the correlation between growth rate, cellular ribosome content and intensity of the hybridization signal [17–19]. As with formazan reduction, introduced by Iturriaga  and Rodriguez et al. , in situ hybridization provides a useful tool for the measurement of individual cell viability.
In the present study, we investigated metabolic potential as well as phylogenetic affiliation of single, adherent bacterial cells during colonization and biofilm formation in drinking water by a combination of physiological and molecular in situ methods.
2Materials and methods
Two modified Robbins devices  were installed as upflow reactors connected to water taps of the domestic drinking water distribution system at the Technical University Berlin. The water taps differed in the frequency of utilization, one had been used about 30 times a day for 5 months before the start of the experiments, the other had been used for 5 months at most. The flow rate during all experiments was 6.0 l h−1, the mean retention time of water in both Robbins devices was 23 min. Physical, chemical and microbiological parameters of the drinking water were analyzed according to standard procedures  and are summarized in Table 1. The Robbins devices consisted of stainless steel cylinders (180 by 150 mm) with 10 threaded holes (diameter 30 mm) each. Low-density polyethylene and glass slides (2×1 cm) were sterilized by UV irradiation (254 nm) for 30 minutes, mounted on the front of stainless steel screws and aligned parallel to the water flow. Slides could be removed independently after various exposure times.
Table 1. Physical, chemical and microbiological parameters of drinking water (Berlin, Jungfernheide)
Means (min.; max.) for 1994
12.2 (9.0; 15.2)
7.4 (7.3; 7.6)
Total hardness (°dH)
18.7 (17.8; 19.7)
KMnO4− O2 (mg l−1)
3 (2; 3)
Total organic carbon (mg l−1)
3.3 (2.8; 3.9)
Ca2+ (mg l−1)
113 (108; 120)
Mg2+ (mg l−1)
12.3 (10.8; 13.7)
Na+ (mg l−1)
57 (34; 77)
K+ (mg l−1)
8.1 (6.0; 10.0)
Fetotal (mg l−1)
0.83 (0.42; 1.09)
Mntotal (mg l−1)
0.50 (0.40; 0.58)
Cl− (mg l−1)
92 (70; 112)
SO42− (mg l−1)
176 (160; 190)
F− (mg l−1)
0.21 (0.19; 0.23)
Cl2 (mg l−1)
Total cell counts (cells ml−1)
By use of a newly developed mechanical mounting system, slides could be placed by insertion into a 3 mm cleft on top of the steel screws and were fixed by a stainless-steel plate (Fig. 1). This mechanical mounting device prevents any vaporization and leaching of chemical substances from organic mounting materials.
2.2In situ hybridization
Fixation of samples and in situ hybridization of adherent biofilm communities were performed according to the protocol described by Manz et al. . A total of 70 biofilms on polyethylene and glass slides were fixed and 321 hybridizations were performed.
Planktonic cells were collected by microfiltration of 100 ml drinking water samples through polycarbonate membranes (25 mm diameter, 0.2 μm pore size, Millipore GmbH, Eschborn, Germany) on nitrocellulose support membranes (pore size 0.4 μm, Millipore GmbH, Eschborn, Germany) using a vacuum filtration unit (Schleicher und Schuell GmbH, Dassel, Germany). All subsequent steps were performed directly in the filtration device, solutions were removed through the membrane by low vacuum pressure. The filtered samples were fixed on the membranes by overlaying them with 0.2 μm pore size filtered, freshly prepared 4% paraformaldehyde solution (Sigma, Deisenhofen, Germany) for at least 1 h at 4°C, washed twice with phosphate buffered saline (PBS, 130 mM NaCl, 7 mM Na2HPO4, 3 mM NaH2PO4, pH 7.2) and subsequently dehydrated with increasing concentrations of ethanol (50%, 80%, and 96%, 3 min each). The polycarbonate membranes were mounted on a glass slide with 15 μl of hybridization solution, overlaid with additional 50 μl of hybridization solution and incubated for at least 1.5 h in a humid chamber at 46°C. Hybridization solution consisted of 0.9 M NaCl, 20 mM Tris/HCl (pH 7.4), 0.01% SDS, 20% (v/v) formamide for probes EUB 338, EUK516, ALF1b, LEG705, and SRB385, 35% (v/v) formamide for probes BET42a, CF319a, and GAM42a.
Finally the membranes were washed for 15 min in 48°C prewarmed washing solution (20 mM Tris/HCl, 0.01% SDS, containing 88 mM NaCl for probes BET42a, CF319a, GAM42a, and 250 mM NaCl for probe EUB338, EUK516, ALF1b, LEG705, SRB385), and mounted on a glass slide using antifading reagent (Citifluor AF2, Citifluor Ltd., London, UK).
The following oligonucleotides were used: (i) EUK516, specific for the domain Eucarya ; (ii) EUB338, complementary to a region of the 16S rRNA conserved in the domain Bacteria ; non-EUB338, complementary to EUB338, serving as a negative control for nonspecific binding; (iii) ALF1b, complementary to a region of the 16S rRNA characteristic for the α-subclass of Proteobacteria ; (iv) BET42a and GAM42a, oligonucleotides complementary to a region of the 23S rRNA of the β- (BET42a) and γ- (GAM42a) subclass of Proteobacteria ; (v) SRB385 complementary to a region of the 16S rRNA conserved in members of the δ-subclass of Proteobacteria, including sulfate-reducing bacteria and some gram-positive bacteria ; (vi) CF319a, specific for the flavobacteria-cytophaga group of the phylum cytophaga-flavobacter-bacteroides ; and (vii) LEG705, specific for the family Legionellaceae . The oligonucleotides were purchased from TIB MOLBIOL, Berlin, Germany, and labeled with tetramethylrhodamine-5-isothiocyanate (TRITC; Molecular Probes, Eugene, USA) or with 5(6)-carboxy-fluorescein-N-hydrosuccinimide ester (FLUOS; Boehringer Mannheim, Mannheim, Germany) as described previously .
2.4Total cell counts
Cell counts of planktonic and surface associated bacteria were determined according to the protocol of Hicks et al.  for dual staining of samples with 4′,6-diamidino-2-phenylindole (DAPI) and fluorescence-labeled rRNA probes. The protocol was modified and staining was performed for 5 min at a final concentration of 1 μg of DAPI ml−1 after hybridization and washing. For determination of total cell counts of planktonic cells, 50 ml samples of drinking water were concentrated by microfiltration through a polycarbonate membrane filter (Millipore, 0.2 μm pore size) and stained with DAPI at a final concentration of 1 μg ml−1.
2.5Heterotrophic plate counts
Biofilms were removed from the Robbins device after different exposure times and immediately placed in sterile drinking water (filtered through a 0.2 μm nitrocellulose membrane). Bacteria were detached from the slide surface with a sterile plastic scraper and pooled in a total volume of 2 ml sterile drinking water. To determine the total cell count, 0.5 ml aliquots of the suspensions were filtered through a 0.2 μm polycarbonate membrane and stained with DAPI as described above. Serial dilutions of the obtained bacterial suspensions were plated on R2A agar  and plate counts were enumerated after 7 days of incubation at 21°C.
Biofilms were removed from the Robbins device after different exposure times and immediately placed in sterile drinking water or in 0.5×R2A medium, both containing 0.5 mM 5-cyano-2,3-ditolyl tetrazolium chloride (CTC, Polysciences Inc., Eppelheim, Germany). The biofilm was incubated at room temperature in the dark for 2 h, rinsed in distilled water and fixed with formaldehyde solution (4% v/v). DAPI staining was performed as described above and the biofilm was mounted in Citifluor. CTC solutions were prepared immediately before use.
2.7Probe active counts (PAC)
The protocol for determination of direct viable counts proposed by Kogure et al.  was modified by use of pipemidic acid (analytical grade, Sigma, Deisenhofen, Germany) at a final concentration of 10 mg l−1 in 30 ml 0.5×R2A medium followed by incubation of biofilms at 21°C in the dark for 8 h. After incubation, biofilms were fixed in 4% formaldehyde solution (v/v) for 2 h, washed once in phosphate buffered saline (PBS, 130 mM NaCl, 7 mM Na2HPO4, 3 mM NaH2PO4, pH 7.2) and dried at 21°C.
Instead of the enumeration of elongated cells, probe active counts were determined after in situ hybridization using the Bacteria-specific probe EUB338 and microscopic enumeration of fluorescent cells. To determine the amount of cells detached from the biofilm during incubation, the medium was filtered through a polycarbonate membrane filter (0.2 μm pore size) and the cells were stained with DAPI as described above. The number of detached bacteria per cm2 of biofilm surface (D) was calculated using the following equation:
where n is the number of bacteria per cm2 after filtration of the assay medium on the membrane, Am is the membrane surface area and Ab is the surface area of the incubated biofilm.
The performance of the gyrase inhibitors pipemidic acid, nalidixic acid and piromidic acid (analytical grade, Sigma, Deisenhofen, Germany) was evaluated in parallel assays by comparing cell counts on biofilms incubated in 0.5×R2A in the presence of one of the antibiotics with cell counts on the biofilm prior to incubation. Stock solutions of antibiotics were prepared at concentrations of 10 g l−1 in 0.05 M NaOH and stored at −20°C. Cell counts were determined after staining with DAPI as described above.
2.8Microscopy and documentation
Fluorescence was detected by epifluorescence microscopy with a Zeiss Axioskop (Oberkochen, Germany) fitted with a 50W high-pressure bulb and Zeiss light filter set no. 01 for DAPI (excitation 365 nm, dichroic mirror 395 nm, suppression 397 nm), no. 09 for FLUOS (excitation 450–490 nm, dichroic mirror 510 nm, suppression 520 nm), and no. 15 for CTC (excitation 546 nm, dichroic mirror 580 nm, suppression 590 nm). Color micrographs were taken on Kodak EES 1600 color reversal film, exposure times were 8–30 s. For statistical evaluation, at least 10 microscopic fields (100×100 μm) and a minimum of 1000 cells were chosen randomly and enumerated. Statistical analysis (standard error, Student's t-test) were done with SigmaPlot 2.0 (Jandel Scientific, Erkrath, Germany) software programs.
3.1Population densities during primary surface colonization
Both polyethylene and glass slides were readily colonized by bacteria, but total cell counts of population densities varied significantly between the two materials (Fig. 2). When polyethylene was used as substratum, total bacterial cell counts increased rapidly and the maximum cell density of 5.1×106 (±3.3×105) cells per cm2 was reached after 7 days of exposure. On glass surfaces, total cell counts remained significantly lower (P<0.0004) and the maximal bacterial density of 1×106 (±1.5×105) cells per cm2 was reached after 14 days. After the initial colonization of the substrata, cell counts declined concomitantly with the appearance of protozoa within the biofilm community (Fig. 2 and Fig. 3A). Bacterial cell density of biofilm communities obtained on glass surfaces stabilized at mean values of 5.9×105 cells per cm2. In contrast to this, bacterial cell density on polyethylene slides ranged from 7.5×105 to 4.2×106 cells per cm2 and a stable plateau of cell numbers was not reached during the experiment. Protozoa within the biofilm communities were detected and enumerated using the hybridization probe specific for the domain Eucarya (EUK 516). After 7 days of exposure, protozoa could be detected in all biofilms examined and occurred mainly as single cells. In mature biofilms, however, protozoa appeared in tight cell clusters forming characteristic grazing fronts within the bacterial community, and biofilm areas behind such grazing fronts were void of bacteria (Fig. 3A).
3.2Phylogenetic diversity and influence of substratum properties
During all stages of biofilm development, bacterial communities on both substrata were dominated by characteristic rod-shaped bacteria belonging to the β-subclass of Proteobacteria (Fig. 3B). The bacterial diversity of the attached community on glass surfaces was higher than on polyethylene during the early stages of biofilm development, represented by significant amounts of cells belonging to the α- and γ-subclass of Proteobacteria (Fig. 4). This difference, however, disapeared when surfaces were exposed for more than 40 days.
Bacteria hybridizing with the probe specific for members of the α-subclass of Proteobacteria (ALF1b) were present in all stages of biofilm formation on glass and reached a maximum of 14% of the total cell counts after 21 days of exposure. Using polyethylene slides, bacteria belonging to the α-subclass of Proteobacteria were observed in significant amounts (9%) after 42 days of exposure and were mostly located in areas with a very sparsely populated biofilm caused by the grazing activity of protozoa. Proteobacteria belonging to the γ-subclass were represented in the biofilm population with a maximum of 7% of the total bacterial population on glass surfaces and a maximum of 4% on polyethylene after 42 and 70 days of exposure, respectively (Fig. 4). Bacteria belonging to the flavobacteria-cytophaga group could be visualized only in mature biofilms, but cell numbers never reached more than one percent of the total bacterial population (Table 2Table 3). Members of the δ-subclass of Proteobacteria were not detected within any of the biofilm populations using probe SRB385. Grazing activity and uptake of bacteria by protozoa was demonstrated through simultaneous hybridization with the oligonucleotide probes EUB338 and EUK516 (Fig. 3A). Interestingly, bacteria belonging to the family Legionellaceae could be identified with the oligonucleotide probe LEG705. These bacteria were almost exclusively associated with protozoan host cells and displayed a very strong hybridization signal, indicating high metabolic activity of the cells (Fig. 3C). Although only a small fraction of the protozoa in the biofilm were infected with bacteria belonging to the Legionellaceae, they could be detected in most biofilms on polyethylene and glass slides exposed for more than 28 days, and occasionally, single free cells with positive hybridization signals were discovered in the biofilm.
Table 2. Activation of attached bacteria on glass surfaces
35 Days of exposure
56 Days of exposure
70 Days of exposure
aMean values (given in cells per cm2)±standard error.
bAll values are percent of total cell counts determined by DAPI staining.
In situ hybridizationb
Table 3. Activation of attached bacteria on polyethylen surfaces
35 Days of exposure
56 Days of exposure
70 Days of exposure
aMean values (given in cells per cm2)±standard error.
bAll values are percent of total cell counts determined by DAPI staining.
In situ hybridizationb
3.3Phylogenetic diversity of planktonic cells
Planktonic cells were hybridized directly on polycarbonate membranes after filtration of drinking water from the domestic water distribution system. Only a small fraction 23% (±4.2%) of the total bacterial population could be detected by in situ hybridization with the eubacterial probe EUB338. Bacteria belonging to the β-subclass of Proteobacteria accounted for 15% (±2.7%) of the bacterial population and bacteria belonging to the α- and γ-subclasses made up 4% (±4%) and 1% (±1%), respectively. Bacteria belonging to δ-subclass of Proteobacteria or to the flavobacteria-cytophaga group could not be detected among planktonic cells.
3.4Influence of water retention time
To evaluate the influence of the water retention time in the distribution system on the dynamics of the bacterial biofilm population, a Robbins device was connected to a cold water tap which had not been used for several months. Colonization and initial biofilm formation during the first days of exposure was performed almost exclusively by typical coccoid bacteria belonging to the γ-subclass of Proteobacteria; 83% of total cell counts showed strong fluorescence signals after in situ hybridization with the probe GAM42a. This first step in biofilm formation was followed by a rapid population shift of the bacterial community; within 9 days of biofilm development, bacteria hybridizing with probe BET42a increased to 70% of the total population, whereas the formerly dominating population of γ-subclass bacteria dropped to 21%. The succession of different phylogenetic bacterial groups during colonization are shown in Fig. 5. After 21 days of exposure, bacteria belonging to the γ-subclass had almost disappeared from the biofilm, whereas bacteria belonging to the β-subclass still made up 70% of the total bacterial population. The absolute numbers of bacteria belonging to the γ-subclass of Proteobacteria slightly increased from 1.1×105 (±2×104) cells per cm2 after 2 days to 1.3×105 (±6×104) cells per cm2 after 9 days, but decreased thereafter to only 0.8×104 (±0.3×104) cells per cm2 after 21 days of exposure.
In contrast to biofilm formation on polyethylene slides with water from a frequently used tap, bacteria belonging to the α-subclass of Proteobacteria were present during all stages of biofilm formation and increased slightly with longer exposure times (Fig. 4A, Fig. 5).
3.5In situ measurement of metabolic potential
To elucidate changes in the metabolic potential of bacteria during the process of biofilm formation, respiratory activity by CTC reduction and fluorescence signals after in situ hybridization of single bacterial cells were monitored as independent parameters during the first 70 days of biofilm formation in a Robbins device connected to a frequently used tap. The time course of bacterial metabolic potential assessed by CTC reduction and in situ hybridization is given in Fig. 6.
Bacterial respiratory activity was visualized by intracellular reduction of the tetrazolium salt CTC to red fluorescent, water-insoluble formazan crystals . In our system, a concentration of 0.5 mM CTC was sufficient to detect the maximum number of respiring bacteria and an increase of the dye concentration to 5 mM did not result in higher cell numbers.
Assay conditions strongly influenced the amount of CTC-reducing bacteria. As shown in Fig. 6, incubation of biofilms in sterile drinking water resulted in significantly lower percentages of actively respiring cells compared to incubation in carbon amended medium (P<0.0034). Within biofilms grown for 2 days on glass surfaces, 51% (±4%) of total cell counts reduced CTC after incubation in sterile drinking water. A significant increase of actively respiring cells to 86% (±2%) was reached by incubation in 0.5×R2A medium (Fig. 3D). The number of formazan containing, fluorescent cells declined continuously during the first 21 days of slide exposure and stabilized thereafter. The amount of in situ detectable, respiring cells after incubation in sterile drinking water reached a plateau of 2% (±1%) to 5% (±2%) of total cell counts, while the number of respiring cells after activation in 0.5×R2A medium remained between 22% (±5%) and 30% (±7%) of total cell counts (Fig. 6). CTC reduction by biofilm bacteria on polyethylene as substratum was monitored in a parallel experiment. The amount of CTC-reducing biofilm bacteria was comparable to the percentage of respiring cells on glass surfaces during the experiment. The number of respiring cells stabilized at 3% in drinking water and at 30% after incubation in 0.5×R2A.
The percentage of cells containing sufficient ribosomes to yield a detectable hybridization signal was determined using the Bacteria-specific oligonucleotide probe EUB338. In contrast to the number of CTC-reducing cells, the amount of cells with positive hybridization signals declined very slowly during the first 21 days of biofilm formation on glass surfaces from 90% (±3%) to 82% (±2%) (Fig. 6). After 35 days of exposure, the percentage of detectable cells after hybridization was reduced to 52% (±5%) of total cell counts and remained stable at this level. This decrease of fluorescence signal intensity was observed in all experiments performed on glass and polyethylene slides, as shown in Fig. 4A,B.
3.6Heterotrophic plate counts
Plate counts of biofilm bacteria on R2A agar were determined after 2, 21 and 70 days of slide exposure and compared with total cell counts. Plate counts remained almost constant during the early stages of biofilm formation with 0.1% (±0.025%) of total cell counts after 2 days and 0.07% (±0.04%) after 21 days of exposure, but decreased to only 0.01% (±0.001%) after 70 days of exposure.
3.7Probe active counts (PAC)
About 50% of single bacterial cells in mature biofilms could be easily detected by in situ hybridization using fluorescence-labeled oligonucleotide probes. To obtain further information about the remaining 50% of the microbial population, a new method for the assessment of these apparently inactive bacteria was established. In a modification of the direct viable count technique (DVC) proposed by Kogure et al. [27, 28] biofilms were incubated in 0.5×R2A medium containing the gyrase inhibitor pipemidic acid to prevent cell division during activation. Subsequent enumeration of elongated cells was replaced by fluorescent cell counting after in situ hybridization. The amount of bacteria with a sufficient ribosome content to yield a clear fluorescence signal after in situ hybridization with the Bacteria-specific oligonucleotide probe EUB338 was defined as probe active counts (PAC). For evaluation of unspecific staining and autofluorescent cells, control hybridizations with labeled probe non-EUB338 were performed.
R2A medium has been developed for cultivation of bacteria from potable water  and therefore seemed to be most suitable for our experiments. A comparison between incubation of biofilms in 0.5×R2A medium and in yeast extract, as proposed by Kogure et al. [27, 28], revealed a slight difference in the percentage of activated bacteria between the two media. Incubation in yeast extract yielded 79% (±2%) probe active counts and incubation in R2A resulted in 84% (±1%, P=0.015).
The DNA synthesis inhibiting antibiotics nalidixic aid, pipemidic acid and piromidic acid were compared for their efficacy within drinking water biofilms. The initial cell density of 1.2×106 (±9×104) remained nearly stable upon incubation with pipemidic acid (1.1×106, ±1×105, P=0.44). Nalidixic acid and piromidic acid did not effectively suppress cell division, cell numbers increased significantly from 1.2×106 (±9×104) to 2.2×106 (±1.8×105) (P<0.001) and 2.8×106 (±1.7×105) (P<0.001), respectively.
The performance of pipemidic acid was monitored in all activation experiments by enumeration of total cell counts of surface associated bacteria with the fluorochrome DAPI before and after activation. Additionally, the mean values of cells detached from the examined biofilm during incubation were determined by microfiltration of the assay medium and staining of cells with DAPI. Total cell numbers did not change significantly in any of the activation experiments (P>0.18), clearly indicating the effective suppression of bacterial cell division by pipemidic acid (Tables 2 and 3). The amount of detached cells during activation varied between 2% and 6% of total bacterial counts (surface associated and detached cells).
Results of biofilm activation experiments after 35, 56 and 70 days of slide exposure are summarized in Table 2 (glass) and Table 3 (polyethylene). In general, the amount of cells yielding a bright hybridization signal with the probe EUB338 increased from about 50% to more than 80% in activation experiments, with the only exception of a 70-day-old biofilm community grown on glass, which showed a comparably low activation rate with 63% hybridized cells of total cell counts.
Group-specific oligonucleotide probes were used to characterize and compare the bacterial population composition of biofilms before and after activation. The relative amount of bacteria belonging to the β- or γ-subclass of Proteobacteria did not differ significantly between activated and non-activated biofilms on glass or polyethylene slides. Bacteria belonging to the flavobacteria-cytophaga group were present at very low numbers in mature, non-activated biofilms. After activation however, 1% to 7% of the total bacterial population of the biofilm could be identified as members of this taxon. The percentage of bacteria belonging to the α-subclass of Proteobacteria increased in some biofilm populations drastically after activation (e.g., from 4% to 26% in a 70-day-old biofilm on polyethylene and from 5% to 15% in a 56-day-old biofilm on glass), but remained at the same level in other biofilms, showing that subpopulations of mostly inactive bacteria might be present at some stages of biofilm development.
Current knowledge on bacterial biofilms includes information on the initial adhesion of bacteria to surfaces [29, 30] and on structure and dynamics of mature, steady-state biofilms [31, 32] but very little information is available on the gap in between, the process of biofilm formation. In the present study, in situ methods were combined to describe the development of (i) bacterial density, (ii) phylogenetic diversity and (iii) bacterial metabolic potential during the formation of drinking water biofilms.
Two different materials, polyethylene and glass, were used as biofilm substrata in our experiments. Various types of plastic material are widely used in domestic drinking water distribution systems. Polyethylene has been described in a number of studies as hydrophobic material enhancing bacterial attachment and growth [33, 34]; glass slides were used because of the inert nature and hydrophilicity of the material. Bacterial densities on polyethylene were found to be 2 to 10 times higher than on glass, which correlates well with the observation that most bacterial strains isolated from freshwater attach preferentially to hydrophobic surfaces . Additionally, surface roughness and additives leaching from polyethylene might influence the bacterial density of the biofilms . Regardless of the substratum, grazing eukaryotic organisms controlled cell densities after the initial rapid colonization by bacteria. The effect of grazing protozoa on the bacterial biofilm community was demonstrated in situ using oligonucleotide probes specific for the domains Eucarya and Bacteria. These findings are in good agreement with results obtained by Pedersen , who reports a similar grazing effect exerted by protozoa on bacteria in drinking water biofilms.
Using group-specific rRNA-targeted fluorescent oligonucleotide probes to characterize the phylogenetic composition of bacterial communities during biofilm formation, rod-shaped bacteria belonging to the β-subclass of Proteobacteria were found to constitute the predominant group of bacteria in all biofilm communities, independent of the substratum used or the age of the biofilm. The prevalence of members of the β-subclass of Proteobacteria has been reported from other, quite diverse aquatic habitats, such as activated sludge  and oligotrophic aquatic systems  and might be a common feature of most aquatic environments, which has been underestimated using conventional cultivation methods.
In contrast to this general feature we observed a different colonization pattern in the phylogenetic composition of the biofilm community using water from a dead end line. In this case, coccoid bacteria belonging to the γ-subclass of Proteobacteria dominated the initial stages of surface colonization. However, these cells could not establish and form a stable biofilm, but were rapidly displaced by bacteria belonging to the predominant β-subclass bacteria. The absolute numbers of bacteria belonging to the γ-subclass decreased by more than 10-fold during this process, indicating that the cells actively detached from the surface. Such detachment processes might be very important for the dynamics of biofilm development and be partly responsible for population shifts. Our results suggest a strong influence of the water retention time on bacterial population composition. Similar results were reported earlier by LeChevalier and coworkers  based on culture-dependent methods.
Biofilm communities on glass surfaces differed from biofilms on polyethylene by a higher phylogenetic diversity, represented by bacteria belonging to the α- and γ-subclass of Proteobacteria. The question whether this difference is influenced by different physico-chemical surface properties of the substrata or by lower bacterial densities, resulting in less competition between individual members of the bacterial populations, remains to be answered.
Keeping in mind that only one quarter of the whole planktonic bacterial community could be hybridized with oligonucleotide probes, a typical 15:4:1 ratio of hybridized cells belonging to the β-, α-, and γ-subclass of Proteobacteria was observed. The comparison of the phylogenetic composition between planktonic bacteria and young biofilms revealed that β-subclass Proteobacteria might attach more easily to surfaces than members of other bacterial groups and therefore dominate the early stages of biofilm formation.
Bacteria belonging to the Legionella cluster appeared repeatedly in mature biofilms, probably correlated to the transmission by protozoa. The Legionella-like organisms could be almost exclusively detected in association with protozoa, supporting the hypothesis that growth of these bacteria might be restricted to intracellular propagation in their natural environments . The clear fluorescence signal of the intracellular bacteria is a strong indication for their high ribosome content.
Respiratory activity and fluorescence signals after in situ hybridization were determined as independent parameters of bacterial metabolic potential during the first 70 days of biofilm formation. The percentage of CTC reducing and hybridization positive cells within the biofilm community and the number of heterotrophic plate counts were highest during primary surface colonization, declined during the process of biofilm formation and finally reached stable levels after 35 days of experiments.
Actively respiring cells can be visualized by the intracellular reduction of CTC to the red fluorescent CTC-formazan crystal , a method which is used to enumerate respiring cells in different environmental samples [38–42]. In our study, we found a vast difference between the amount of respiring bacteria under in situ conditions and after incubation with nutrients, indicating that a remarkable fraction of cells reduced CTC only in the presence of nutrients. Whereas Rodriguez et al.  and Smith et al.  described the same effect, others reported incubation without nutrients to be as effective [38, 40]. These discrepancies might be due to differences in the affinity of bacterial cells obtained from various habitats to CTC and nutrients, as well as to variations of the incubation protocol . However, the amount of respiring cells declined in both assays to 3% (in situ) and 26% (with nutrient addition) of the total bacterial population after 21 days of slide exposure.
Early studies on the correlation of bacterial ribosome content and growth rate of Salmonella typhimurium have been adopted for conclusions about the metabolic activity of individual cells by measuring fluorescence intensities after whole cell hybridizations . Poulsen and coworkers also demonstrated that decreasing hybridization signals were correlated with the rRNA content of the bacteria and not with cell permeability. In our study, the proportion of bacteria containing a sufficient amount of ribosomes to yield positive hybridization signals declined slowly during the first 21 days of slide exposure from 90% to 82% of the total population and stabilized at 50% after 35 days. In contrast to this, the percentage of planktonic bacteria with a clear hybridization signal was only 23% of the total population, demonstrating the difference in ribosome content between surface-attached and free-water phase bacteria, supporting earlier observations by Manz et al. .
Comparison between these two independent methods showed that fluorescence signals and the supposed corresponding cellular ribosome content declined more slowly than bacterial in situ or substrate-dependent determined respiratory activity. These findings are in agreement with results from Fukui et al. , who suggests a certain amount of a durable fraction of rRNAs in starvation competent bacterial cells.
In drinking water, as in most natural or man-made habitats, more than 90% of the total bacterial population could not be enumerated by heterotrophic plate counts, due to unfavorable culture conditions or the formation of viable but non-culturable states . A number of in situ methods aim to circumvent these known biases by evaluating bacterial viable or active counts.
Most surveys of bacterial direct viable counts [38, 39, 45, 46] are based on the original protocol of Kogure et al. , involving incubation of bacteria with yeast extract in the presence of the gyrase inhibitor nalidixic acid and subsequent determination of elongated cells. In 1984, Kogure  proposed the combined use of three antibiotics, nalidixic acid, piromidic acid and pipemidic acid, but could not report a significant increase in efficacy compared to the incubation with nalidixic as sole inhibitor.
Concerning drinking water biofilms, pipemidic acid was most effective for inhibition of cell division during the 8 h incubation period. In contrast to pipemidic acid, biofilms incubated with nalidixic acid or piromidic acid showed a marked increase of cell numbers during incubation. Although all three antibiotics inhibit DNA synthesis and prevent cell division in a similar manner, pipemidic acid differs from the other two antibiotics in its effectiveness against Pseudomonas species [27, 47, 48], which appeared to be of great importance in drinking water.
We therefore modified the incubation protocol for direct viable counts proposed by Kogure et al. [27, 28] and, instead of enumerating elongated cells which proved to be quite difficult within the heterogeneous biofilm community, we present here a method for the determination of viable cells, PAC (probe active counts), on the basis of fluorescence signals. In situ hybridization using a Bacteria-specific oligonucleotide probe revealed an increase of hybridized cells from 50% to more than 80% of the total bacterial population through activation of the adherent bacteria with appropriate substrates for oligotrophic freshwater habitats. This effect could be repeatedly shown with bacteria grown on different substrata and after different exposure times, which clearly demonstrates the rapid, substrate-dependent reactivation of apparently inactive cells. Bacteria belonging to the flavobacteria-cytophaga group and to the α-subclass of Proteobacteria were seriously underestimated by in situ hybridization in some mature biofilms but could be successfully visualized performing the PAC method. This new method might therefore be an important improvement for rRNA-targeted in situ probing of bacteria in extremely oligotrophic habitats.
We would like to thank Per Ericsson and Cornelis Reitz for their essential help in design and construction of the Robbins devices and the new mechanical slide mounting system.