Microscope-Based Imaging Platform for Large-Scale Analysis of Oral Biofilms

Abstract

A microscopic method for noninvasively monitoring oral biofilms at the macroscale was developed to describe the spatial distribution of biofilms of different bacterial composition on bovine enamel surfaces (BES). For this purpose, oral biofilm was grown in situ on BES that were fixed at approximal sites of individual upper jaw acrylic devices worn by a volunteer for 3 or 5 days. Eubacteria, Streptococcus spp., and Fusobacterium nucleatum were stained using specific fluorescence in situ hybridization (FISH) probes. The resulting fluorescence signals were subsequently tested by confocal laser scanning microscopy (CLSM) and monitored by an automated wide-field microscope-based imaging platform (Scan∧R). Automated image processing and data analysis were conducted by microscope-associated software and followed by statistical evaluation of the results. The full segmentation of biofilm images revealed a random distribution of bacteria across the entire area of the enamel surfaces examined. Significant differences in the composition of the microflora were recorded across individual as well as between different enamel surfaces varying from sparsely colonized (47.26%) after 3 days to almost full surface coverage (84.45%) after 5 days. The enamel plates that were positioned at the back or in the middle of the oral cavity were found to be more suitable for the examination of biofilms up to 3 days old. In conclusion, automated microscopy combined with the use of FISH can enable the efficient visualization and meaningful quantification of bacterial composition over the entire sample surface. Due to the possibility of automation, Scan∧R overcomes the technical limitations of conventional CLSM.

INTRODUCTION

A common phenomenon in nature is the formation of dynamic surface-adherent bacterial structures called biofilms (18, 33). In the oral cavity in particular, multispecies oral biofilms consist of more than 700 species of bacteria, enmeshed in a polysaccharide-rich extracellular matrix (15, 25, 40). These specialized microbial biofilms have evolved to endure in the adverse environment of tooth surfaces and gingival epithelium (26, 32). Due to this, their development is the outcome of numerous complex physicochemical interactions between oral-tissue substrata, microorganisms, and adsorbed macromolecules (16, 48). The initial colonization of oral biofilms is characterized by the adherence of oral streptococci to pellicle proteins such as alpha-amylase, proline-rich proteins, and glycoproteins or to host receptors on epithelial cells (28, 43). Further oral biofilm formation can then be attributed either to coadhesion resulting from recognition between planktonic cells and surface-bound cells or to coaggregation due to cell-cell recognition between different species of bacteria (26). Fusobacterium nucleatum coordinates the binding of early and late colonizers, particularly obligate anaerobes and streptococci (21). In addition, bacteria interact with polysaccharides, mostly glucans, and other salivary proteins, contributing to the maturation of dental plaque biofilms (23).

The biofilm research community has been showing an increasing interest in detecting and understanding the mechanisms of biofilm formation over the past decade. This microbiology-oriented focus was promoted by the fact that dental plaque biofilms have been demonstrated to be crucial etiological factors in biofilm-mediated diseases such as caries, gingivitis, and periodontal disease (11, 31, 46). The thin biofilm layer of Gram-positive microorganisms that characterizes healthy patients is altered into a chunky heterogeneous structure under the prevalence of numerous Gram-negative bacteria, usually observed in dental infections (27, 49). Complex interspecies interactions, including coaggregation, bacteriocin production, and metabolic and quorum-sensing communication among oral bacteria, are notable features of this multispecies biofilm community. Bacterial biofilm communities also present up to 1,000-times-higher resistance to antimicrobial agents, host immunity, and variations in other environmental factors, such as pH and oxygen, than do their planktonic counterparts (13, 47, 52). Therefore, the elimination of these persistent well-structured bacterial ecosystems could play an integral role in finding efficient prevention and competent treatment strategies against oral diseases.

The development of new methodologies for the visualization of in situ- or in vivo-established biofilms has introduced a new era in this field of dental research (10, 20). Among the traditional methods used, viable plate count and culture-dependent techniques may inadvertently select for certain species in multispecies biofilms and can only poorly quantify semiplanktonic or desorbed biofilms (29). High-resolution microscopic techniques such as scanning electron microscopy (SEM), environmental scanning electron microscopy (ESEM), transmission electron microscopy (TEM), cryo-electron microscopy, and atomic force microscopy (AFM) make the ultrastructure of bacteria and their environment visible, but they cannot quantify adherent microorganisms (19). Among the more current bacterial identification techniques, the fluorescence in situ hybridization (FISH) method allows for the in situ study of the spatial and temporal dynamics of the bacterial community by means of fluorescently labeled oligonucleotide probes (5). The combined use of FISH with confocal laser scanning microscopy (CLSM) provides high-resolution three-dimensional (3D) images of individual microbial members in their natural environment, but only over a limited sampling of the total surface area (1). Thus, CLSM monitors a specific confocal field, without taking into consideration the extensive discrepancies among various fields within the same biofilm (22). Optical absorption and scattering also constrain the depth of optical penetration of CLSM, whereas unlabeled compounds and other spatial structure features cannot be easily detected (44).

In view of the methodical restrictions involved in the established biofilm visualization techniques mentioned above, image acquisition in our lab was done for the first time over an entire sample surface using a modular microscope-based screening station (Scan∧R; Olympus Europa, Hamburg, Germany). We attempted to show how the Scan∧R imaging platform based on high-content-screening microscope technology can fully automate image processing and data analysis of high-content large-scale images, achieving a full segmentation of the image into subareas with evident biofilm formation. This report aims to describe and refine the characteristics of an experimental approach for the identification of oral bacteria in intact in situ-grown biofilms by the use of a Scan∧R imaging platform combined with FISH. For this purpose, oral biofilms grown in vivo were examined for the presence of Fusobacterium nucleatum and Streptococcus spp. using species-specific FISH probes in addition to a general eubacterial probe after 3 and 5 days, respectively. The results demonstrate that this visualization method is a promising technique for representative analysis of oral biofilm composition.

MATERIALS AND METHODS

Subjects and specimens.

A healthy 33-year-old volunteer participated in the study. Thorough clinical examination by an experienced dentist revealed good oral hygiene status. The volunteer abstained from antibiotics 6 months prior to the beginning of the study. The study design was reviewed and approved by the Ethics Committee of the University of Freiburg (proposal 222/08).

For the preparation of the specimens, the buccal surfaces of bovine incisors of 2-year-old cattle were removed and modified into cylindrical enamel specimens (diameter, 5 mm; 19.63-mm² surface area; height, 1.5 mm). Their bovine spongiform encephalopathy (BSE)-free status was confirmed after examination with the IDEXX Laboratories BSE diagnostic kit (Ludwigsburg, Germany). The enamel surfaces of all samples were then polished by wet grinding with abrasive paper (400 to 4,000 grit). The protocol for disinfection of the enamel plates included ultrasonication in NaOCl (3%) for 3 min to remove the superficial smear layer, air drying, and ultrasonication in 70% ethanol for another 3 min. The disinfected samples were then ultrasonicated twice in double-distilled water for 10 min and, finally, stored in distilled water for 24 h to hydrate prior to exposure in the oral cavity (1).

Red wax was used to fix the enamel specimens on the approximal sides of an individual upper-jaw acrylic appliance over periods of 3 and 5 days. Six disinfected enamel slabs were placed in the interdental area between upper premolars and molars, so that the movements of the tongue or cheek could not inhibit biofilm formation (Fig. 1). The visualization and quantification of the adherent bacteria were achieved by combined use of FISH with the Scan∧R screening microscope (2). For each time period, the volunteer carried six enamel samples that were rinsed off with sterile 0.9% saline solution for 10 s after their exposure in the oral cavity.

Fig 1.

Individual upper-jaw acrylic appliance with the enamel slabs in place at different locations. The specimens were positioned at the front (f), in the middle (m), and at the back (b), on both sides, right (R) and left (L), of the appliance. The exposed surfaces were fixed toward the tooth enamel by red wax.

FISH.

FISH was carried out according to a modified protocol for the enamel slabs, previously described by Amann et al. and modified by Al-Ahmad et al. (3, 5). In brief, after removal from the oral cavity, the biofilms formed on the enamel slabs were fixed in 4% paraformaldehyde in phosphate-buffered saline (PBS; 1.7 mM KH₂PO₄, 5 mM Na₂HPO₄ with 0.15 M sodium chloride [pH 7.2]) for 16 h at 4°C. The specimens were then rinsed off twice with sterile PBS and fixed again in a solution composed of ethanol (50%) in PBS (1:1, vol/vol) for 12 h at 4°C. The probes were subsequently washed twice with PBS and permeabilized in a solution comprising 7 mg/ml of lysozyme (hen egg white lysozyme; Fluka, Buchs, Switzerland; 105,000 U/mg in 0.1 M Tris-HCl, 5 mM EDTA [pH 7.2]) for 9 min at 37°C in a humid chamber. Thereafter, the specimens were rinsed off twice with PBS and dehydrated in an ascending series of ethanol concentrations (50 to 100%) for 3 min each.

The hybridization of the enamel slabs was conducted in 24-well plates (Greiner Bio-One, Frickenhausen, Germany), wrapped with aluminum foil, at 46°C for 90 min. Each well contained 600 μl hybridization buffer (5 M NaCl, 1 M Tris-HCl [pH 8.0], 25% [vol/vol] formamide, and 10% [wt/vol] sodium dodecyl sulfate [SDS]) and 1 μl oligonucleotide probe for eubacteria, fusobacteria, and streptococci (50 ng/μl). The hybridization of the specimens followed their incubation in 600 μl wash buffer (159 mM NaCl, 20 mM Tris-HCl [pH 7.5], 5 mM EDTA [pH 8.0], and 0.01% [wt/vol] SDS) per well for 15 min at 48°C.

High-pressure liquid chromatography (HPLC)-purified oligonucleotide probes for eubacteria, Streptococcus spp., and F. nucleatum were commercially synthesized and 5′-end labeled (Thermo Fisher Scientific GmbH, Ulm, Germany). The specificity of the oligonucleotide probes (EUB 338, FUS 664, and STR 405) listed in Table 1 was tested previously using different bacterial strains (3).

Table 1.

Sequences, 5′ modifications, target species, and references for used oligonucleotide probes

Probe	Sequence (5′–3′)	5′ modification	Target	Reference
EUB 338	GCTGCCTCCCGTAGGAGT	Fluorescein	Eubacteria	4
FUS 664	CTTGTAGTTCCGC(C/T)TACCTC	Cy5	F. nucleatum	50
STR 405	TAGCCGTCCCTTTCTGGT	Cy3	Streptococcus spp.	42

CLSM.

All FISH probes were excited at the following wavelengths: fluorescein, 488 nm; Cy3, 546 nm; and Cy5, 633 nm. The measurement of the fluorescence emission of the probes was carried out at the following wavelengths: fluorescein, 495 to 565 nm; Cy3, 552 to 592 nm; and Cy5, 649 to 703 nm. Further analysis of the labeled biofilms was conducted by CLSM (Leica TCS SP2 AOBS) by placing the specimens face down onto a drop of physiological saline solution in a chambered coverslip (μ-Slide, 8 well; Ibidi, Germany) and examining them with the aid of a 63× water immersion objective (HCX PL APO/bd.BL 63.0 × 1.2 W; Leica). Sequential scanning was used to minimize the risk of spectral overlap between the probes (2). In brief, three representative locations on the enamel plates were chosen for the visualization of the oral biofilm. The upper and lower boundaries of the biofilm at each location were determined, and the mean thickness of the biofilm was calculated using these measurements from the three locations. Biofilms were scanned in the Z-direction at these three points, producing sections of a thickness of approximately 0.5 μm each at 2-μm intervals throughout the biofilm layers. Each image was taken with a resolution of 1,024 by 1,024 pixels and a zoom setting of 1.7, corresponding to physical dimensions of 140 by 140 μm.

Automated microscopy. (i) Sample preparation and instrument calibration.

After the in situ hybridization of the enamel probes, the specimens were covered in a mixture of 20 μl Peakflow TM beads (Blue flow cytometry reference beads; Invitrogen, Germany) and PBS (1:10 [vol/10 vol]). The samples were then left to dry in the dark for 1 h at room temperature. The blue beads were detected using a 4′,6-diamidino-2-phenylindole (DAPI) filter set (excitation, 352 nm to 402 nm; emission, 437 nm to 475 nm). These beads were used to enable autofocusing of the microscope, whether or not biofilm was present at a particular location (Fig. 2). The samples were then placed face down in a chambered coverslip as for the confocal microscopy experiments.

Fig 2.

Cross-section of oral biofilm on bovine enamel slabs. Blue reference beads were applied on the biofilm surface for instrument setup and calibration.

(ii) Image acquisition.

Image acquisition was conducted using a 20× objective (UPLSAPO N.A. 0.75) on a Scan∧R high-content-screening system based on an IX-81 inverse microscope stage (Olympus Europe, Hamburg, Germany). The excitation of the probes was carried out with the MT20 illumination system (xenon-mercury burner, 150 mW; Olympus, Germany). For measurement of the FISH fluorescence, the following filter sets were used: fluorescein isothiocyanate (FITC) (excitation, 494 to 504 nm; emission, 534 to 549 nm), Cy3 (excitation, 540 to 550 nm; emission, 583 to 594 nm), and Cy5 (excitation, 620 to 650 nm; emission, 700 to 737 nm).

In order to determine the location of the outer boundaries of the samples, three representative locations were chosen along the edge of each enamel slab with the aid of the Scan∧R software. Each of the 12 slabs was scanned at up to 177 nonoverlapping positions, covering the entire surface of the plate, in such a way that each image was completely contained within the area of the plate. Each of these images was derived from Z-stacks made up of 10 sections taken at intervals of 2 μm and constituting a total imaging depth of 20 μm. Because we were interested in the plate biofilm coverage rather than the three-dimensional (3D) structure of the biofilm, these images were then combined as maximal projections at each of the positions and for each of the channels.

(iii) Image analysis.

Each image consisted of 1,346 by 1,024 pixels and covered a total area of 433 by 330 μm. Within the Scan∧R analysis program, every image in turn was subdivided into 475 virtual subobjects of 50 by 50 pixels or approximately 16 by 16 μm each; these were then used as the smallest units of measurement for the subsequent data analysis. Therefore, each enamel slab had a total number of up to 84,075 data points which could be used for measurement. The goal was to try and look globally at the sample, rather than capture individual bacteria. The thresholding was carried out on these virtual subobjects through the use of dot plots depicting mean FITC signal in the Scan∧R software and then enclosed within a region, similar to that used in typical flow cytometry analyses. The subobjects which were contained within this region were considered to be positive for the eubacterial probe and were controlled again by referencing them to the original images. The number of subobjects was then used to calculate the total amount of subobjects containing eubacteria. Similarly, the mean intensities of Cy3 and Cy5 were also used to determine whether or not a region was positive for these probes; however, in both of these cases, the subobjects had to also be positive for the FITC signal in order to minimize the occurrence of any possible false-positive subobjects from the Cy3 and Cy5 channels. A background correction was performed for each channel, and the presence or absence of biofilm was determined by setting intensity thresholds for each of the three probes (eubacteria, Streptococcus spp., and F. nucleatum), which were then used across the entire sample. The evaluated regions with positive fluorescence signals for eubacteria (FITC) were used to determine the 100% value of the bacterial biofilm on the plate and therefore to determine the coverage level of the samples in percentages. For a subobject to be considered positive for Streptococcus spp. or F. nucleatum, it had to also be positive for eubacteria. The area covered by these target bacterial cells was then measured as a percentage of the FITC-positive regions.

Statistical analysis.

For the descriptive exploration of the data, box plots were calculated and graphically displayed, stratified by side, position, and time. The coefficient of variation was calculated stratified by side, position, and time. An analysis of variance (ANOVA) was used. For each examined subject (eubacteria, Streptococcus spp., and F. nucleatum), the continuous response variable was modeled as a function of side, position, and the corresponding interaction as explanatory variables, and separately for time. Model assumptions were graphically checked by residuals and other regression diagnostics (including Cook's distance). The normality of error terms was assumed. All calculations were done using PROC MIXED from the statistical software SAS 9.1.2.

RESULTS

CLSM.

Representative images of biofilm formation are presented in Fig. 3. After FISH, the visualization of the bacterial biofilms on the enamel surfaces was tested by confocal laser scanning microscopy (CLSM). All bacterial species examined were detectable on the enamel slabs after exposure in the oral cavity for 3 days. The morphology of the majority of the microorganisms was coccoid, but filamentous bacteria of various configurations were also identified. Inter- and intraindividual variety was manifested upon observation of single aggregates, mono- or multilayered bacterial chains, and three-dimensional bacterial clumps of various sizes.

Fig 3.

FISH-CLSM micrographs of enamel slabs after 3-day in situ exposure in the oral cavity to control the specificity of FISH probes. Green, eubacteria (a); yellow, F. nucleatum (b); magenta, Streptococcus spp. (c).

Automated microscopy. (i) Quantitative analysis of biofilm formation after 3 days.

Figure 4 shows a FISH-Scan∧R tiled image paired with the graphical distributions of the different regions which were positive for the respective bacterial targets (eubacteria, Streptococcus spp., and F. nucleatum) of a 3-day-old oral biofilm grown in situ on enamel slabs.

Fig 4.

FISH-Scan∧R image and graphical distribution of different biofilm bacteria (eubacteria, Streptococcus spp., and F. nucleatum) of a 3-day-old oral biofilm grown in situ on enamel slabs. The image seen in panel A is a tiled overview of the microbial colonization pattern showing an overlay of all three staining patterns. Oral biofilm was stained simultaneously with the all-bacterium-specific EUB 338 probe (green), the Streptococcus-specific EUB 405 probe (red), and the F. nucleatum-specific FUS 664 probe (blue). Panels B, C, and D graphically depict the spatial distribution of positive subobjects after thresholding and analysis in the Scan∧R analysis program, containing eubacteria, Streptococcus spp., and F. nucleatum and labeled with FITC, Cy3, and Cy5, respectively. x and y axis values represent the different positions on the microscope stage.

Three days after biofilm formation, 47.26% of the enamel surface was coated with bacteria. The covering grades of the substratum by Streptococcus spp. and Fusobacterium nucleatum were 19.73% and 3.45% after 3 days, respectively. Statistical analysis showed significant differences (P ≤ 0.0337) in the contents of eubacteria and Streptococcus spp. on the enamel surfaces located on the left (31.38% and 17.71%) and right (63.15% and 21.24%) sides of the oral cavity, respectively. Differences between the two sides were not significant for F. nucleatum (P = 0.1752). Additionally, the bacterial content seemed to vary depending on the position of the enamel slabs in the oral cavity. Less biofilm was formed on the specimens situated in the middle (25.55%) than on the front (53.04%) and posterior (63.15%) specimens (P < 0.0.0001). The posterior samples also had higher contents of Streptococcus spp. and F. nucleatum (42.5% and 6.87%) than the front (11.06% and 2.05%) and middle (5.1% and 1.12%) specimens, respectively (P < 0.0001).

(ii) Quantitative analysis of biofilm formation after 5 days.

In Fig. 5, tiled overviews of images derived from the scanning of biofilm-coated enamel plates hybridized with three different fluorescence probes are presented. The quantitative results of the bacteria detected are also shown in Fig. 6 in the form of box plots.

Fig 5.

Overviews of tiled images constructed from the individual Scan∧R maximum-projection images of the supporting enamel surface of representative 5-day-old in situ oral biofilms on bovine enamel plates. Oral biofilm was hybridized with three different specific probes (EUB 338, FUS 664, and STR 405). Shown are results obtained with green eubacterium-specific probe (EUB 338) (a), red Streptococcus-specific probe (STR 405) (b), and blue Fusobacterium nucleatum-specific probe (FUS 664) (c) after 5 days and all three probes after 5 days (d). Scale bars, 20 μm.

Fig 6.

Box plots depicting percentages of the different bacterial members in 3- and 5-day-old oral biofilm as detected by FISH. The central line is the median; whiskers indicate minimum and maximum. Each box indicates the front (f), middle (m), or back (b) position of the enamel slabs on the right (r) and left (l) sides of the oral cavity, respectively.

Five days after biofilm formation, 84.45% of the enamel surface was covered with microorganisms. Streptococcus spp. and Fusobacterium nucleatum showed covering grades of 49.37% and 15.98% after 5 days, respectively. Statistical differences (P ≤ 0.00003) in the coverage of eubacteria and Streptococcus spp. were detected between the left (87.59% and 55.16%) and the right (81.59% and 43.58%) specimens in the oral cavity, respectively. Differences between the two sides were not significant for F. nucleatum (P = 0.8309). As far as the bacterial content in different positions of the samples in the oral cavity is concerned, a higher percentage of eubacteria was detected on the front (87.64%) specimens than on the middle (82.8%) and on the posterior (83.31%) samples (P < 0.0332). The posterior samples also had a higher content of Streptococcus spp. (56.65%) than those positioned at the front (47.83%) or in the middle (43.63%) of the oral cavity (P < 0.0001). F. nucleatum was mainly located on the middle (20.28%) and posterior (20.66%) surfaces and less on the front (7%) ones (P < 0.0001). The percentages presented above are all median values.

DISCUSSION

For the first time, the current study has established a minimally destructive scanning approach to examine oral biofilm composition across entire sample surfaces by the combined use of FISH with 16S rRNA-targeted oligonucleotide probes and automated microscopy. This enabled an overall assessment of oral bacterial diversity and distribution across the enamel slabs. The perception of biofilms as landscapes is one of the latest trends in oral biofilm research, focusing on the large-scale distribution and movements of microorganisms in the framework of microbial biogeography (8). Indeed, understanding the role of microorganisms in oral diseases such as dental caries, gingivitis, and periodontitis necessitates a thorough visualization of the spatial distribution of a predominant species (7, 54). In addition to this, a clear vision of the entire biofilm structure unravels the underlying mechanisms of spatial biofilm patterning in relation to biofilm development. Numerous factors which affect biofilm formation and development, such as complex interspecies interactions, differences in local substrate concentrations or other nutrients, and environmental changes (pH and oxygen), can be highlighted through large-scale image observation (8). In their study, Nielsen et al. investigated a flow cell-grown model consisting of Burkholderia sp. strain LB 400 and Pseudomonas sp. strain B13 (FR1) which interacted metabolically (37). With the aid of the FISH-CLSM technique, they successfully monitored a shift in the spatial structure of the consortium, after a shift from noncommensal to commensal conditions. Furthermore, the outcomes of antimicrobial treatment on oral biofilms often involve structural alterations accompanied by functional changes within the biofilms. In a study by Hope and Wilson, the influences of 0.05 and 0.2% (wt/vol) chlorhexidine (CHX) were visualized by CLSM on multispecies biofilms, modeling interproximal plaque grown on a hydroxyapatite substrate (24). Image analysis revealed biofilm contraction at a rate of 1.176 μm⁻¹ after exposure to 0.2% CHX.

The use of bovine enamel slabs provides a representative model for the study of bacterial colonization on surfaces. A great number of easily gained, homogeneous bovine enamel surfaces were used in this study due to their common physicochemical properties with human dental hard tissues. Past studies have indicated that there are no detectable differences between the compositions of the biofilms formed on artificial and natural tooth surfaces (38), whereas there are some evident discrepancies in the composition of the pellicle (39).

A parameter of great relevance, the covering grade of the entire enamel surface, has been examined in this study. The number of microorganisms on the enamel slabs increased significantly with the oral exposure time. Thus, the total proportions of Streptococcus spp. and F. nucleatum were 23.18% and 65.35% of the total bacteria after 3 and 5 days, respectively. This corresponds roughly to the values for vital bacteria as reported by Arweiler et al. (6). Coaggregation plays a key role in bacterial colonization of the tooth surface (29). This is especially important for Streptococcus spp. and F. nucleatum. FISH studies conducted at time intervals of longer than 72 h may reveal interactions between the different bacterial species (41). A CLSM study by Dige et al. demonstrated the predominance of streptococci in biofilm during the first 6 to 48 h (14). Streptococci secrete the extracellular enzyme glycosyltransferase. In saliva and in the pellicle, three isoforms of this enzyme are usually detected: glycosyltransferases B, C, and D (51). These enzymes synthesize specific receptors for the attachment of streptococci, which have a relatively short doubling time (9). The above factors could offer an explanation for the dominance of streptococci in the first days of dental plaque maturation. However, the presence of streptococci might be overemphasized because the culture methods used may select for certain species (3).

Fusobacterium nucleatum, a Gram-negative microorganism, has the ability to coaggregate with early and late colonizers (45). Therefore, F. nucleatum has an extremely important function as a bridge among oral bacteria. It is considered to be a relevant initiator of periodontal disease resulting from matured subgingival plaque. Guggenheim et al. studied a biofilm formed in vitro and consisting of five species (17). After 2.5 days, F. nucleatum was detected at a percentage of 50%. In our study, F. nucleatum represented only a small proportion of the initially adhering bacteria that was influenced by the exposure time of the enamel slabs. After 5 days, F. nucleatum represented a greater part (15.98%) of the adherent bacteria than after 3 days (3.45%), and it was mainly situated on the middle and posterior samples. A possible explanation could involve proband-specific characteristics and the presence of higher oxygen concentrations at the frontal sites of the oral cavity, since the presence of the late colonizer F. nucleatum is bound to low oxygen concentration (17).

The results of the present study showed some very significant differences in the local distribution of Streptococcus spp. and F. nucleatum among the different specimens. The enamel slabs that were positioned at the back or in the middle of the oral cavity have proven to be more suitable for the examination of biofilms up to 3 days old. In contrast, the front enamel slabs would be indicated for studies in which biofilms older than 5 days are involved. Previous reports showed no influence of the position of the specimens on the data (3). However, the current results are more representative due to the greater number of measuring points obtained with the automated microscope. Further interpretation implies differences in oxygen supply of the samples, the availability of saliva, and the saliva flow rate, as well as differences in tooth anatomy.

There is currently an increasing demand for proper quantification of the highly variable phenotypes found in oral biofilms. The subjectivity due to a low number of representative images can be avoided only by conducting a higher number of objective measurements of biofilm composition. Therefore, automated microscopy incorporates automated image acquisition of multistained oral bacteria and image analysis of these data facilitating automated scanning of entire surfaces. The automated determination of the focal plane for each field of view is a crucial parameter for acquiring stable and reproducible biofilm images. Low-quality screenings of irrelevant biofilm components with intense fluorescence signal or of only a small number of oral microorganisms usually result from focusing either on the entire field of view or on the most eye-catching plane containing visible structures (30). For this reason, we decided to use fluorescent microbeads added to the sample after biofilm formation to identify the most appropriate focal plane, independently of whether bacterial biofilm was present or not at a given location. The use of a 20× objective was deemed to be sufficient for our purposes. It allowed the acquisition of images covering the entire area of interest on the enamel slabs with enough detail to be informative, but at the same time allowing the scans to be completed within a reasonable amount of time and without a drying out of the substrate. Another critical factor concerning image acquisition is the autofluorescence of the substrate. After a number of pilot studies in which the background of enamel autofluorescence was minimal, we carried out our experiments with natural tooth surfaces, despite the fact that in some instances autofluorescence has been shown to impede the interpretation of fluorescence signals of in situ biofilms (53).

In order to obtain robust large-scale images, efficient staining of the different oral microbial species and reproducible image acquisition were essential prerequisites. The bacterial populations of interest constituting the biofilms could be quantified and compared after the application of fluorescence in situ hybridization (FISH). The FISH protocol applied in this study, in which the oligonucleotide probes were labeled with FITC, Cy3, or Cy5, delivered specific fluorescence signals from eubacteria, Streptococcus spp., or F. nucleatum, respectively, and the signals could be easily distinguished after merging images from the different color channels. Some of the most important advantages of FISH for the identification of bacteria in oral biofilms are the ability to detect uncultured bacteria and the rapid availability of new oligonucleotide probes. However, the restrictions of this method are caused by the fact that the specific oligonucleotide FISH probes can stain only intact ribosomes in vital bacteria. Ribosomes in dead bacterial cells are degraded quickly due to loss of membrane integrity (5). Finally, stable illumination and sensitive detection of the emitted fluorescence during the excitation of the oligonucleotide probes were taken into consideration as supplementary factors affecting the perceived image quality (35).

For the automated image analysis after background correction, the positive regions for each of the fluorescent probes were identified via a combination of thresholding based on fluorescence intensity and gating. These thresholds were applied to all of the images from a single plate. To our knowledge, there has been no software solution to overcome the technical restrictions concerning automatic recognition of highly diverse bacterial populations. Some open-source software packages such as CellProfiler Analyst/Cell Classifier and EBImage constitute fast image-processing engines capable of classification by machine learning of cell recognition, but they are not optimal for use with the irregular structures normally encountered in biofilms (12). The development of new software specialized in the detection of bacterial cells and biofilm structures could automate the control of a whole range of fluorescence-based imaging assays in microbiology, since high-content-screening platforms such as Scan∧R are computer controlled and fully motorized (12).

The scanning of a single enamel slab (5 mm in diameter) as described in this report took approximately 2 h to complete. At first glance, this could be considered too time-consuming, especially compared with relatively faster visualization techniques such as CLSM. However, the ability of automated microscopy to measure large-scale biofilm development across an entire surface area with high spatial resolution has its own advantages. In the confocal microscopy experiments, 3 positions were chosen “randomly” within a plate, with each position covering a region of 140 μm by 140 μm. In contrast, the Scan∧R autofocus allowed “hands-off” scanning, which was able to follow surface irregularities across an entire plate, in addition to the ability to tile the positions of the image stacks. The circular shape of the plates allowed us to use a “3-point method” to set the boundaries for each plate, enabling the entire surface area of the plate to be scanned automatically. Scanning an entire plate brings a level of objectivity that is difficult to obtain when choosing positions by hand, however carefully it is done. The automated microscopy approach also allowed analysis of the entire plate as a unit, rather than having to examine multiple individual stacks. Covering a much larger surface than is practical to do manually allowed us to gather much more information per plate, leading to improved statistical analysis.

Magnetic resonance imaging (MRI) would appear to be another option for noninvasive visualization of bacterial structures with various applications on dental biofilms. Mapping of structural biofilm characteristics, biofilm structure/flow relation, and flow and transport properties as well as diffusion profiles constitutes information demonstrated in MRI time-resolved studies. In a recent study, minimally destructive monitoring of biofilm processes was conducted by combining an integrated nuclear magnetic resonance (NMR) system with a confocal laser scanning microscope. The entire three-dimensional microscopic structure of the biofilm was successfully visualized by MRI, while complementary fluorescence information obtained by CLSM increased our knowledge about oral biofilms in a depth-resolved fashion (34). At first glance, MRI appears to be suitable for mapping the spatial characteristics of most kinds of oral biofilms. Although this may be true for in vitro-grown monospecies biofilms thicker than 200 μm, a challenge is posed by the significantly thinner in situ-grown multispecies oral biofilms. The technical restrictions of this method are highlighted by the lack of contrast in the density of ¹H nuclei between the microbial biomass and the bulk phase, especially if we consider that most oral biofilms consist of more than 95% water (36). Unless additional contrast is applied to the images, they are of poor quality and unsuitable for further analysis. The relatively low resolution of the device—approximately 7 to 15 μm—is definitely a factor contributing to the failure of this technique to efficiently scan and monitor dental biofilms thinner than 200 μm. On the other hand, automated microscopy offers higher resolution, which allows full mapping of the locations of the bacterial communities present and, combined with FISH detection, enables the specific and simultaneous identification of multiple bacterial species within the biofilm.

From an ecosystematic point of view, variations in oral biofilms can be detected only if a representative area of the biofilm is analyzed. The automated microscopy-based system used in this study allows the minimally destructive investigation of intact biofilm across the complete surface of the enamel plate. The Scan∧R has the ability to produce and further analyze multiple thin sections of the biofilm (30). In contrast, the CLSM technique manages to analyze only a restricted amount of data due to a limited number of available measuring points. The Scan∧R software enables the quantification of various bacterial populations and their metabolic activities within the oral biofilm. The measured fluorescence intensity then correlates with the metabolic activity of the bacteria within the biofilm. Alterations in bacterial metabolic activity could be caused by different dental materials, nutrient components, and antibacterial mouthwashes.

To our knowledge, there have been no studies to date which have conducted large screenings of oral biofilm composition. Only the study of McLean et al. (34) managed to visualize in vitro-grown single-species biofilm using nuclear magnetic resonance and confocal microscopy. However, this technique is not appropriate for distinguishing different bacterial species within a multispecies biofilm. Moreover, the automated microscopy technique presented here has already been used for the visualization of large numbers of eukaryotic cells (12). In the present study, automated microscopy proved to be an efficient method for the screening of large-scale in situ-grown oral biofilms. Until now, the existing reports have described visualization techniques with the aid of confocal microscopy to examine representative areas of the oral biofilm. Future studies should compare this automated microscopy-based method with the conventional confocal microscopy method which has been used to date.

In conclusion, fluorescence labeling techniques such as FISH are applicable to the evaluation of bacterial adherence to enamel surfaces. Therefore, the screening ability of automated microscopy, combined with the use of FISH, enables the efficient visualization and meaningful quantification of bacterial populations across an entire sample surface. In the future, the same methodological approach should be applied to more individuals in order to investigate more bacterial species, as well as to study biofilm growth on other materials such as composites, fiberglass posts, and implant materials.