Joel D. Rudney, Department of Diagnostic and Biological Sciences, University of Minnesota School of Dentistry, 17-252 Moos Tower, 515 Delaware St. SE, Minneapolis, MN 55455, USA. E-mail: email@example.com
Most studies of biofilm effects on dental materials use single-species biofilms, or consortia. Microcosm biofilms grown directly from saliva or plaque are much more diverse, but difficult to characterize. We used the Human Oral Microbial Identification Microarray (HOMIM) to validate a reproducible oral microcosm model.
Methods and Results
Saliva and dental plaque were collected from adults and children. Hydroxyapatite and dental composite discs were inoculated with either saliva or plaque, and microcosm biofilms were grown in a CDC biofilm reactor. In later experiments, the reactor was pulsed with sucrose. DNA from inoculums and microcosms was analysed by HOMIM for 272 species. Microcosms included about 60% of species from the original inoculum. Biofilms grown on hydroxyapatite and composites were extremely similar. Sucrose pulsing decreased diversity and pH, but increased the abundance of Streptococcus and Veillonella. Biofilms from the same donor, grown at different times, clustered together.
This model produced reproducible microcosm biofilms that were representative of the oral microbiota. Sucrose induced changes associated with dental caries.
Significance and Impact of the Study
This is the first use of HOMIM to validate an oral microcosm model that can be used to study the effects of complex biofilms on dental materials.
Composite resin materials have become an increasingly popular choice for restoring carious teeth, due to their superior aesthetic qualities relative to dental amalgam. However, composite restorations fail more often than amalgam restorations, and this is primarily due to the development of secondary caries around the margin of composite restorations (Soncini et al. 2007). Recent years have seen increased interest in potential interactions between dental composite materials and oral biofilms. Composite restorations can provide a substrate for attachment of oral biofilms, and it is important to know whether the quantity or composition of adherent biofilms differs among composite materials according to their physical, chemical and material properties. By the same token, it is also important to know whether biofilms exert effects on the lifespan of composite restorations in the mouth, either directly, by degrading the composite, or indirectly, by attacking the interface between composites and oral tissues. The long-term goal of such studies is to develop materials that retain desirable physical properties, while limiting biofilm growth and extending the lifetime of restorations in the mouth (Busscher et al. 2010).
A variety of in vitro models have been proposed for investigating such questions. Designs differ, but one major factor that may influence the generalizability of results is the nature of the biofilms that are used. Many investigators have chosen to simplify and standardize their models by using single-species biofilms. Streptococcus mutans is often chosen (Beyth et al. 2008; Fucio et al. 2008, 2009; Gyo et al. 2008; Kantorski et al. 2008; Buergers et al. 2009). Many oral streptococci can produce lactic acid as a by-product of sucrose metabolism (Sansone et al. 1993). However, S. mutans is particularly acidogenic, and S. mutans-mediated acid demineralization of enamel and dentin has generally been considered as the primary aetiologic agent for dental caries (Smith and Spatafora 2012). The major limitation of S. mutans biofilm models is that single-species biofilms do not exist in the mouth, where S. mutans can be a minority species even in persons with active caries (Aas et al. 2008).
As an alternative, models have been developed that employ consortia of up to nine species, incorporating mixtures of commensal species with putative pathogens (Dezelic et al. 2009; Rinastiti et al. 2010; Hayati et al. 2011). However, with the current diversity of the oral flora reported to be >700 species (Dewhirst et al. 2010), consortia still may be an unrepresentative sample of the species that materials may be exposed to in the mouth. Some have attempted to address that problem by mounting materials in appliances that are carried intraorally (Auschill et al. 2002; Sousa et al. 2009). However, in situ studies can take a long time to carry out, and the appliances can be uncomfortable for participants.
Yet another alternative is to use an oral inoculum to generate a complex diverse microcosm biofilm in vitro. Most typically, saliva from a single donor or pooled donors has been used to provide the inoculum. Microcosm models can be manipulated in various ways to simulate aspects of oral conditions such as the flow of saliva and dietary carbohydrates across biofilm surfaces (Sissons et al. 1991, 1995, 2007; Pratten et al. 1998, 2000; Matharu et al. 2001; Badawi et al. 2003; Mcbain et al. 2005; Ledder et al. 2006; Filoche et al. 2007; Cenci et al. 2009).
One limitation of microcosm models has been that, because of their complexity, microcosms have been difficult to standardize or characterize. Fortunately, advances in culture-free methods for microbial identification have provided a way to address this problem. In later studies, denaturing gradient gel electrophoresis (DGGE) or checkerboard DNA–DNA hybridization has been used to identify up to 40 species in oral microcosms (Rasiah et al. 2005; Ledder et al. 2006; Filoche et al. 2007; Sissons et al. 2007). Now, the recently available Human Oral Microbial Identification Microarray (HOMIM) system has expanded the range of the checkerboard assay almost sevenfold, by using reverse-capture ribosomal RNA probes to detect and evaluate the relative abundance of 272 distinct oral taxa (Colombo et al. 2009).
Since its introduction, the HOMIM system has been widely used in clinical studies of the microbiology of oral diseases (Colombo et al. 2009, 2012; Ahn et al. 2011; Lif Holgerson et al. 2011; Olson et al. 2011; Tanner et al. 2011; Luo et al. 2012). However, it has not yet been used to characterize in vitro multispecies models of oral biofilm. Our purpose in this study was to use the HOMIM system as a tool for validating and establishing the reproducibility of an oral microcosm model based on the commercially available CDC Biofilm Reactor, which we are using to investigate biofilm interactions with composite resins at restoration interfaces. We addressed the following questions: (i) Is the composition of microbial inoculums obtained from the same donors stable over time? (ii) How similar are saliva and plaque inoculums obtained from the same donor at different times? (iii) How reproducible are microcosms grown from saliva or plaque inoculums taken from the same donors at different times? (iv) Does the species composition of saliva and plaque microcosms differ from their original inoculums? (v) How are microcosm composition and pH affected by pulsing sucrose into the biofilm reactor environment? (vi) How similar are microcosms formed on different types of substrates within the same reactor? (vii) Can reproducible microcosms be grown from frozen stocks of previous microcosms?
Materials and methods
Questions 1–5 and 7 above were addressed using saliva and dental plaque inoculums obtained from four adult subjects, all of whom were authors of this manuscript. All four were in good general health, and none had taken antibiotics within the past 3 months. Three subjects were in good oral health, as determined by the study dentist at the time of initial sampling. However, one subject was found to have periodontal disease. We decided to retain the data from that person in the analyses described below, reasoning that his/her clinical status was not likely to bias our ability to validate our oral microcosm model.
A different and larger population of subjects was obtained for the second phase of studies to address Questions 5–7. That group consisted of a convenience sample of 10 paediatric patients already participating in a larger ongoing study using HOMIM to compare oral biofilms associated with the margins of amalgam and composite restorations. Each child was examined by the study paediatric dentist, who made a formal caries risk assessment (CAMBRA) (Ramos-Gomez et al. 2010), which found them to be at high risk for future caries. Two children had active carious lesions at the time of sampling for this study. All children were otherwise in good general health and had not taken antibiotics within 3 months of saliva and plaque sampling. Their average age was 8·5 years.
Resting whole saliva was collected by expectoration. The study dentist collected dental plaque inoculums from either the occlusal or buccal margin of existing restorations. A sterile sickle scaler was used, and each sample was immediately deposited into a vial containing 1 ml prereduced anaerobic transfer medium (Anaerobe Systems, Morgan Hill, CA, USA). All procedures involving human subjects were approved by the University of Minnesota Institutional Review Board.
Major components of the model
The core component of our oral microcosm model is the CDC Biofilm Reactor (Goeres et al. 2005), which is available from BioSurface Technologies, Bozeman, MT, USA. It incorporates a lidded vessel through which growth media can be flowed at a defined rate, and a baffled stir bar to generate shear. Substrates for biofilm growth are mounted into eight rods (each rod can hold three discs) that can be removed and replaced aseptically through the lid (Fig. 1a). We used basal mucin medium (BMM) as the growth medium (Sissons et al. 2007). This is a complex medium, with hog gastric mucin as the primary source of carbohydrate. BMM has been used successfully in previous oral microcosm models.
We chose to incubate the biofilm reactor aerobically, to better simulate the ecological succession occurring after supragingival tooth surfaces are initially colonized from saliva. Previous consortium studies have shown that obligate anaerobes can survive in aerobic conditions, when facultative species are also present (Bradshaw et al. 1996). We reasoned that anaerobes in saliva and plaque inoculums likewise would initially be protected by facultative partners, and would then be able to grow anaerobically as microcosm biofilm depth increased, and oxygen continued to be consumed by facultative species. For most experiments described below, the CDC reactor was filled with 350 ml of BMM and placed on a digital heating stir plate. One of the rods was replaced by a sealed port for the temperature probe of the stir plate, so that the internal temperature of the reactor vessel could be maintained at 37°C. The stirring rate was set at 125 RPM. After inoculation (see below), the reactor was incubated under conditions of shear, but no media flow, for 24 h. BMM was then pumped through the reactor at an initial flow rate of 17 ml min−1 for 48 h (the position of the exit port maintains a constant volume of approx. 500 ml of media throughout the flow phase) (Fig. 1b). In sucrose pulsing experiments described below, the flow rate was increased to 20 ml min−1 on the second day of flow. The system was taken down after 2 days of flow, and microcosm biofilms were removed for analysis (see below).
Most components of the system were autoclavable, and most components were also cleaned with bleach between uses. Seventy percent ethanol was used to disinfect materials that could not be bleached or autoclaved and to maintain aseptic conditions under circumstances such as sucrose pulsing.
Substrates used for biofilm growth included discs (12 mm in diameter) of hydroxyapatite (HA; Clarkson Chromatography, South Williamsport, PA, USA), and silorane-based composite (LS; 3M, Saint Paul, MN, USA). A methacrylate-based composite (Z100; 3M) was also used in later sucrose pulsing experiments. These discs were sized to fit the sample mounts in the CDC reactor rods (Fig. 1a).
Experiments using saliva as an inoculum
For our initial experiments with this system, we used resting whole saliva collected from the adult subjects. Saliva was coated directly on HA and LS discs, and also added directly to the reactor. There were three discs per material: one for plate counts, one for scanning electron microscopy and one for HOMIM analysis. A portion of each saliva sample also was retained for DNA extraction and HOMIM analysis (see below). A saliva sample was taken from each subject on three different weeks and inoculated into the reactor to evaluate the reproducibility of microcosms grown from saliva taken from the same donor at different times. Samples were not pooled, and a different sample was used each time the reactor was run.
Experiments using plaque as an inoculum
The same four adults were used as plaque donors for these subsequent experiments. Each first provided a sample of resting whole saliva, and then a sample of supragingival plaque was collected as described above. The saliva sample was clarified by centrifugation, diluted twofold in a buffer that simulates the ionic composition of saliva (Gibbons and Etherden 1985) and then sterilized with a 0·2-μm filter. The matching plaque sample was dispersed by sonification, and a portion was retained for DNA extraction and HOMIM analysis. HA and LS discs were precoated with sterilized saliva, to form a pellicle. We then dispensed 30 μl of matched plaque suspension onto each disc and placed the rods into the reactor. A plaque and sterilized saliva sample were taken from each subject on three different weeks, to evaluate the reproducibility of microcosms grown from plaque taken from the same donor at different times.
Experiments using plaque with sucrose pulsing
Sucrose pulsing experiments were carried out exclusively using inoculums from plaque samples. In the first phase of those studies, we used the four adult donors. No sucrose was added to the reactor during the first day (shear without flow). Then on each of the flow days, the reactor was pulsed five times with 20% sucrose (to achieve a final concentration of 5% sucrose in the reactor volume). The first pulse was administered between 8:30 and 9:00 am and repeated every 2 h thereafter. The five-pulse schedule was meant to be analogous to three meals and two snacks, so sucrose pulsing was discontinued during the nights. Frozen stocks were prepared from microcosms grown during the first sucrose-pulsed run for two of the four donors, using the method described below. Stocks from both subjects then were inoculated into the reactor three times (in different weeks) to evaluate the reproducibility of sucrose-pulsed microcosms grown from stocks taken from different donors.
The paediatric population described above was used for the second phase of sucrose-pulsing studies. Matching saliva and plaque samples were collected as above. Each subject was sampled only once. The sucrose-pulsing protocol was carried out as above. However, discs of HA, LS and Z100 were used in the reactor, to look for any differences in biofilm composition on a composite with a more conventional chemistry than silorane-based LS. The composite discs also were polished to a 0·1-μm finish, to provide a more clinically relevant surface. The surface properties of those discs also were evaluated, and the results will be described in a forthcoming article. To minimize observed effects of autoclaving and radiation on the surface properties (Aparicio et al. 2012), the discs were only disinfected with 70% ethanol before use. A preliminary HOMIM comparison of biofilms from reactors with autoclaved and ethanol-wiped discs had indicated that this did not cause any change in microcosm composition within the same donor (Aparicio et al. 2012).
Biofilms were collected from designated discs into 1 ml prereduced anaerobic transfer medium and dispersed by sonification. Tenfold serial dilutions were prepared in sterile phosphate-buffered saline and plated on a Todd–Hewitt broth-based agar medium containing 5% defibrinated sheep blood, 0·05 μg ml−1 hemin and 0·005 μg ml−1 menadione. The plates were incubated anaerobically, to protect anaerobes present in the microcosms from oxygen after they were dispersed and plated. Plates were then counted after 72 h at 37°C.
Scanning electron microscopy
Designated biofilm discs were initially fixed with 2% glutaraldehyde and 0·15% Alcian blue in 0·1 mol l−1 sodium cacodylate buffer and stored at 4°C. They then were fixed with osmium tetraoxide in cacodylate buffer, dehydrated in a graded series of ethanol concentrations from 50 to 100% and then treated with a critical point drier. They were sputter coated with platinum and then viewed with a field-emission gun scanning electron microscope (FE-SEM) (6500; JEOL, Tokyo, Japan) (Erlandsen et al. 2004).
Frozen stocks were made by growing plaque microcosms from individual subjects, on up to 12 HA discs, in the same reactor. The microcosm biofilms were removed, pooled for each subject, re-suspended in BMM with 20% glycerol and stored at −80°C.
Real-time pH measurement
To determine the effect of sucrose pulsing on biofilm pH, a fitting was machined so that an autoclavable pH electrode could be inserted in place of one of the CDC reactor sampling rods. A recording pH meter was used to collect pH information every 15 min throughout the 72-h incubation period, as biofilm formed on the electrode. Frozen stocks made from paediatric subject microcosms were used for those experiments.
DNA was extracted from all samples using the recommended HOMIM protocol (http://mim.forsyth.org/docs/DNA%20Isolation%20Protocol.pdf). This was carried out on the day of collection for saliva and plaque, and at the end of the 72-h incubation, for microcosm samples. DNA extracts were stored at −80°C and periodically shipped to the HOMIM analysis core at the Forsyth Dental Center (Boston, MA, USA). Whenever possible, samples taken from the four adult subjects on different weeks were shipped together, so they could be analysed during the same run. A detailed description of the HOMIM protocol including PCR primers, thermal cycling conditions, labelling, hybridization and normalization has been published previously (Colombo et al. 2009). Briefly, a sample DNA extract was amplified in two separate PCR, using forward and reverse primers providing overlapping coverage of the bacterial 16S rDNA gene. The products from both PCRs were pooled, purified and then labelled by incorporating Cy3-deoxycytidine triphosphate during a third round of amplification. The HOMIM microarray consists of aldehyde-coated glass slides printed with oligonucleotide reverse-capture probes directed towards species and phylotype-specific regions derived from alignments of 16S rRNA sequences (a complete listing of HOMIM probe sequences is provided in the supplementary material to Colombo et al. 2009). Each array includes positive control probes directed towards universal regions of bacterial 16S rRNA, and negative controls to determine array background. There are five replicate arrays on each slide, and each replicate incorporates duplicate subarrays. The purified labelled PCR products were hybridized to the HOMIM arrays under conditions described by Colombo et al. (2009). The HOMIM arrays were then scanned to determine probe fluorescence intensity. The intensities of replicate spots for the same probes were adjusted for background and averaged to obtain probe-specific values. Those values were normalized for interarray comparison relative to the average intensities for the universal 16S rRNA probes and categorized into relative intensity values ranging from 0 to 5 (the minimum threshold for signal detection is equivalent to approx. 104 bacterial cells). A detailed mathematical description of the normalization process is provided online at: http://bioinformatics.forsyth.org/homim/index.php?name=PNphpBB2&file=viewtopic&t=681&sid=96e9923d75f17f55a9e7562bafbfc377. Results were returned to us in the form of a colour-coded spreadsheet, where the rows designated the reverse-capture rRNA probes currently available, and the columns designated the samples. A number from 0 to 5 in each cell provided a semiquantitative estimate of the relative abundance of rDNA within each sample that hybridized with each probe. The relative abundance data were subjected to further statistical analysis as described below.
Hierarchical cluster analyses using average-linkage and Pearson correlations were used to visualize the extent to which microbial composition was similar between samples from different sources (saliva, plaque and their respective microcosms), the same source under different conditions (plaque microcosms with and without sucrose), repeat samples taken from the same subjects at different times and different materials within the same reactor. A mixed effects model was used to test for differences in bacterial relative abundance levels between conditions for all probes on the array. Rather than constructing a model that simply has additive effects for the various conditions, the model was specified in a manner that allowed for arbitrary differences in the species composition between the following five groups: the saliva inoculum samples, the saliva microcosm samples, the plaque inoculum samples, the plaque microcosm samples and sucrose-pulsed plaque microcosms. The experimental factors (e.g. saliva vs plaque) were treated as fixed effects, and subject-specific additive random effects were introduced to account for the correlation of observations coming from the same subject as is customary in the analysis of longitudinal data (restricted maximum likelihood was used to compute parameter estimates).
To test for differences in the species compositions attributable to sample conditions, permutation tests were conducted in a manner that retained the longitudinal structure of the data set, that is, the permutations were constrained so that data were only permuted within a subject (this is an extension of the typical application of permutation tests to paired data, where the set of allowed permutations just randomizes treatment assignment within the pairs of data). To correct for test multiplicity, a Bonferroni adjustment was used where the total number of tests was taken to be the number of probes on the array for which the mixed effects model was able to produce an estimate, which here was 218 probes (the algorithm fails when the variance of the probe intensity across samples is too low). Consequently, 10 000 permutations were generated for each probe (with the model being re-estimated for each permutation). To test for an effect of time, mixed effects models were fit that also included the repeat number as a factor, and a likelihood ratio test statistic was computed to test for an effect of time (P-values were again computed using permutations rather than by comparison with a chi-square distribution). These calculations were performed using R 2·12·2 using the nlme package (http://www.r-project.org).
A complete list of probes that showed significant differences between conditions is provided in Supporting Information as Table S1. As many of the probes on the array hybridize to multiple species and some species have multiple probes (some of which bind multiple species), determination of species that differ between conditions is complicated. Here, we say a species' population differs between conditions if there is a significant difference for all probes for that species and at least one of these probes is specific to that species. Notes in Table S1 provide more detailed descriptions of how some of those decisions were made.
The microcosm protocol used here produced very dense biofilms after 72 h. Disc plate counts ranged from 108 to 1010 CFU ml−1. There were no discernible differences between inoculum sources, the presence or absence of sucrose, or different materials within the same reactor (not shown).
Disc biofilms showed a complex multilayered structure. Different strata were visible, as well as channels, which appeared to penetrate to disc surfaces. Higher magnifications showed a very diverse array of bacterial morphotypes in close association with each other. Direct contact was observed between many different types of cells, and a web-like network of polysaccharide matrix also was present (Fig. 2).
General patterns of species prevalence
Among the four adult subjects, there was a core group of 16 HOMIM probes that produced a positive signal in at least 70% of all DNA extracts regardless of source (saliva or plaque inoculums, saliva or plaque microcosms, or sucrose-pulsed plaque microcosms) (Table 1). Those probes detected members of common oral genera, including Streptococcus, Veillonella, Haemophilus, Fusobacterium, Slackia and Lachnospiraceae. The 10 paediatric subjects likewise showed a core group of 22 probes that were positive in at least 70% of all DNA extracts regardless of source (plaque inoculums or sucrose-pulsed plaque microcosms) (Table 1). There was substantial overlap in genus/species coverage between both core groups, although the paediatric core also included Prevotella and Tannerella species. By contrast, 48% of probes in the adult samples and 55% in the paediatric samples failed to reach the threshold for signal detection in any sample. There likewise was substantial overlap between the groups with regard to the probes that were negative.
Table 1. HOMIM probes that were positive in 70% or more of DNA extracts from all types of samples, including saliva, saliva microcosms, dental plaque and plaque microcosms with and without sucrose pulsing
Streptococcus cristatus and sp clone BM035_ot058_578_AA47
Granulicatella adiacens and elegans_ot534_596_W81
Fusobacterium nucleatum ss nucleatum and animalis_ot420_698_AE01
Saliva inoculums and microcosms
On average 66 ± 21 (mean ± SD) HOMIM probes were positive for saliva inoculums from the adult subjects. Cluster analysis of the HOMIM patterns showed that the saliva inoculums formed a cluster that was distinct from saliva microcosms, plaque inoculums, plaque microcosms and sucrose-pulsed plaque microcosms. Repeat saliva inoculums taken on different weeks from the same subject clustered together (except for Subject 4), and there were no significant effects over time from the mixed model analysis. This suggested that the microbial composition of saliva generally was consistent within subjects over time (Fig. 3).
The average number of positive probes decreased to 42 ± 12 in microcosms grown from the adult subjects' saliva, suggesting that some species were lost from the original inoculums, or else decreased in relative abundance to the point where they were no longer detectable by HOMIM. A number of anaerobes were retained in the microcosms, including Capnocytophaga, Prevotella and Fusobacterium species, so aerobic incubation per se did not appear to be a factor in those changes. The permutation tests indicated that 26 probes decreased significantly in relative abundance, whereas three probes showed significant increases (Table S1). Rothia mucilaginosa and other bacteria recognized by Rothia probes were relatively more abundant in saliva inoculums, but decreased in the microcosms. Others that decreased in the microcosms included Neisseria flavescens, several Eubacterium species, Shuttleworthia satelles and two Gemella species.
Several Streptococcus probes showed statistically significant decreases in relative abundance between saliva and the microcosms, notably Streptococcus salivarius. While there are a total of 34 HOMIM probes that interrogate various species of Streptococcus, only nine of these are unique to a particular species. The rest measure multiple species due to crosshybridization resulting from sequence similarity. Other Streptococcus species that were significantly lower were australis, infantis, mitis bv 2, clone FN042 and oralis. However, those were all from probes that crosshybridize among the various strains. One Streptococcus probe did show an increase in relative abundance in the microcosms. It recognized Streptococcus cristatus, and clone BM035 (Table S1).
Microcosms grown from saliva inoculums taken on different weeks from the same subject clustered together, and there were no significant effects over time from the mixed model analysis, suggesting that saliva-derived microcosms were stable within subjects. Saliva microcosms mostly clustered separately from plaque microcosms, with the exception of Subject 1. Microcosms grown on HAP and LS at the same time from the same inoculum always clustered in pairs, suggesting that those substrates had little effect on saliva microcosm composition (Fig. 3).
Plaque inoculums and microcosms
The average positive probe count for plaque inoculums from the adult subjects (67 ± 12) was almost identical to that of saliva, but plaque formed a cluster that was distinct from saliva, and from microcosms grown from either saliva or plaque. Repeat plaque inoculums taken on different weeks from the same subject generally clustered together, and there were no significant effects over time from the mixed model analysis, suggesting that the microbial composition of plaque was consistent within subjects over time (Fig. 3).
The average number of positive probes decreased to 39 ± 8 in plaque microcosms from the adult subjects. The permutation tests indicated that 41 probes decreased significantly in relative abundance, whereas 10 probes showed significant increases (Table S1). As with the saliva microcosms, probes from multiple Streptococcus species (oralis, mitis bv 2, anginosus and sanguinis) decreased in relative abundance in the plaque microcosms. However, it was difficult to determine specific species due to crosshybridization of the probes. We also found that the following species were higher in the plaque inoculum samples compared with the plaque microcosms: Tannerella clone BU063, Campylobacter gracilis, Kingella oralis, Neisseria elongata, Neisseria flavescens, Actinobaculum clone EL030, Actinomyces gerencseriae, Actinomyces clone AP064 and Propionibacterium propionicum. In addition to these species, six probes that interrogate various species of Selenomonas were found to decrease, but our criteria for detecting a particular species were not fulfilled (Table S1). We also note that while three probes specific for Cardiobacterium hominis detected significant decreases, there was one probe that crosshybridizes to this species and another Cardiobacterium species for which the mixed model could not produce an estimate (Table S1).
The relative abundance of Veillonella parvula was likely higher in the microcosm samples, as the signal strengths for all three probes that were specific to this species were significantly greater than the original inoculums, as was one of the probes that crosshybridized with one other Veillonella species (another of those crosshybridizing probes produced data for which the mixed model could not generate an estimate). The relative abundances of Mycoplasma salivarium, Fusobacterium periodontium and Leptotrichia clone DR011 also were higher in the microcosm samples (Table S1).
Microcosms grown from plaque inoculums taken on different weeks from the same subject generally clustered together although Subjects 1 and 3 were somewhat less consistent than Subjects 2 and 4. There were no significant effects over time from the mixed model analysis. Collectively, this suggested that plaque-derived microcosms were stable within subjects. Microcosms grown on HAP and LS at the same time from the same inoculum always clustered in pairs (with the exception of Repeat 3 for Subject 3), suggesting that those substrates had little effect on saliva microcosm composition (Fig. 3).
Effect of sucrose pulsing on plaque microcosms
Sucrose-pulsed microcosms from adult Subjects 1–4 clustered separately from their counterparts grown without sucrose (Fig. 3), and their average species count decreased to 25 ± 3. The permutation tests indicated that 23 probes decreased significantly in relative abundance with sucrose pulsing, whereas 14 probes showed significant increases (Table S1). Campylobacter showae, Capnocytophaga clone X066, Haemophilus parainfluenzae, Solobacterium moorei, Fusobacterium periodontium, Gemella morbillorum, Lachnospiraceae clone DO016, Leptotrichia clone DR011, Synergistes clone BH017 and Campylobacter concisus all decreased significantly. Both probes that are designed to measure the combination of Fusobacterium species nucleatum subspecies nucleatum and nucleatum subspecies animalis also decreased, as did both probes that interrogate the combination of the species Eikenella corrodens, Kingella denitrificans and Kingella species clone DE012 (Table S1).
As noted above, interpretation of the results for Streptococcus is complicated by crosshybridization of the probes. There is one probe that is specific to Streptococcus anginosus, and four other probes that detect this species in combination with one other species. All of those probes increased upon sucrose pulsing. Other species that may have increased are Streptococcus intermedius and Streptococcus gordonii (for which there are no fully specific probes), and Streptococcus parasanguis (Table S1).
Sucrose-pulsed microcosms grown from frozen stocks on different weeks clustered together with their parent sucrose-pulsed plaque microcosms (stocks were run only for Subjects 1 and 3, due to considerations of cost). There were no significant effects over time from the mixed model analysis. This suggested that reproducible subject-specific sucrose-pulsed microcosms could be grown from frozen stocks, and that it was not necessary to obtain a fresh inoculum every time a microcosm was grown. Microcosms grown on HAP and LS at the same time from the same inoculum always clustered in pairs, suggesting that those substrates had little effect on saliva microcosm composition (Fig. 3).
Sucrose-pulsed microcosms grown from plaque derived from the 10 paediatric donors also clustered separately from their original inoculums (Fig. 4). The patterns of species loss and gain were quite similar to those found for the adult subjects. In addition to the taxa named above, Actinomyces and Prevotella species dropped below the limit of detection, and decreases in the relative abundance of Streptococcus mitis and sanguinis were seen. The same Streptococcus species that increased in the sucrose-pulsed microcosms from adults likewise increased when the inoculums were from children, and an increase in Veillonella atypica also was observed. Despite these donors' history of early caries, S. mutans was only detected in one plaque sample (from one of the two caries-active subjects), and it was not retained in the corresponding sucrose-pulsed microcosm. A Lactobacillus species was only detected in one plaque sample (from the other caries-active subject), and it also was not retained in the microcosm. Z100 was added as a third material in these experiments, and microcosms grown from the same inoculum on HA, LS and Z100 always clustered in triplets, suggesting that those substrates had little effect on saliva microcosm composition (Fig. 4). Frozen stocks from these donors also yielded reproducible microcosms (not shown).
Effect of sucrose pulsing on biofilm pH
When frozen stocks from paediatric subjects were grown without sucrose pulsing, biofilm pH remained relatively stable, generally ranging between 6·5 and 7·0 throughout the 72-h incubation period. By contrast, sucrose pulsing on Days 2 and 3 induced a rapid drop in pH after the first pulse. On each day, pH remained below 5·5 for approx. 10 h, and then rose above 6·0 at night while sucrose pulsing was discontinued (Fig. 5).
One concern with oral microcosm models is that it has been difficult to study their composition, or determine how well they represent the oral flora. The HOMIM technology made it possible for us to address those questions. We were able to characterize microcosms grown from inoculums taken from the same subjects at different times and show that species composition of both the inoculums and the microcosms were stable. This increases confidence that our microcosm model can be used to generate consistent results. Similar findings have been established for other microcosm models, using earlier techniques such as selective culture, DGGE and checkerboard hybridization (Mcbain et al. 2005; Rasiah et al. 2005).
Saliva most often has been used as an inoculum in previous microcosm models. Our data suggest that saliva inoculums are not entirely similar to plaque inoculums obtained from the same subjects, and that each type of inoculum will yield a microcosm, which reflects those differences. That is not surprising, because previous checkerboard hybridization and 16S rDNA sequencing studies have shown that the saliva flora is more similar to that of the tongue than it is to dental plaque (Kazor et al. 2003; Mager et al. 2003). The extent to which this difference presents an issue depends on the experimental goals of a microcosm study. We have chosen to use plaque inoculums in further studies, because our primary interest is modelling interactions at the interface between restorations and tooth surfaces that lead to secondary caries.
It must be noted that the microcosms we obtained did not completely reproduce the flora of either plaque or saliva inoculums. There may be many reasons for this. We chose to incubate aerobically, because our goal was to simulate the ecological succession of supragingival plaque. As anticipated from previous consortium studies (Bradshaw et al. 1996), a number of anaerobes were retained in the microcosms. However, that does not rule out the possibility that some particularly fastidious anaerobes may have been lost. Other species may have been lost because BMM failed to fully satisfy their nutritional needs. An alternative medium that may support the growth of a wider range of species than BMM has been proposed (Tian et al. 2010). However, it did not appear to increase diversity in our system during preliminary studies comparing it to BMM (not shown). Overall, the observation that some species were lost needs to be weighed against the fact that the species with the highest relative abundance in inoculums were retained in the microcosms.
Sucrose pulsing seemed to have a potent effect on microcosm composition. A number of species were lost from the microcosms, while others increased in relative abundance. Our real-time pH measurements suggest that this is most likely due to acidification of the system, although it also is possible that some species are unable to survive in the presence of sucrose. We do not consider the species changes that accompany sucrose pulsing to be a drawback of our model, because such changes may reflect those that occur during the sucrose-mediated ecological shift that leads to dental caries. Interestingly, the lack of change in S. mutans levels that we observed with the sole S. mutans-positive subject has also been seen in other sucrose-pulsed microcosm models (Filoche et al. 2007). It is possible that sucrose pulsing selected for acidogenic nonmutans streptococci, because the pH of sucrose-pulsed microcosms consistently dropped below 5·5 during the pulsing periods. This suggests that microcosm models can be used to model secondary caries, regardless of whether S. mutans is present within the microcosm.
One environmental factor that did not appear to affect microcosm composition in our model was the nature of the substrate inoculated with either saliva or plaque. HA, LS and Z100 biofilms obtained from the same inoculum at the same time consistently clustered very closely. One reason for this might be that planktonic bacteria released from growing biofilms were free to colonize other discs and that would tend to cancel out any initial differences in growth patterns between materials. However, it also is possible that there were no differences from the outset, because the materials exerted no selective effects. Our ongoing studies will involve the use of materials with antimicrobial properties, and that will help to resolve this question. It is important to note that HA, LS and Z100 did show some differences in their surface properties after exposure to biofilms, and that will be the subject of a companion article.
One characteristic of our CDC biofilm reactor model that differs from other published oral microcosm models is that it produces large quantities of biofilm in a relatively short period of time (biofilm depths exceeded 500 μm after 72 h, as determined by crosspolarization optical coherence tomography) (Chen et al. 2012). This can be seen as an advantage or disadvantage. On the one hand, the CDC reactor provides a high throughput system with a capacity for multiple samples, and it appears to speed up processes that take longer to occur in the mouth. That may be an advantage compared with other systems in which an experiment may take 30 or more days to complete. On the other hand, the CDC reactor does not simulate the plaque, saliva and air interface in the way that the constant depth film fermenter and drip flow reactor do, and that may make it less realistic. From our perspective, every model system has its strengths and weaknesses, and results in which the most confidence can be placed will be those which are reproducible in multiple model systems. In that respect, it is encouraging that patterns of intraindividual reproducibility, interindividual variation and species changes with sucrose pulsing similar to those we observed have been seen with other microcosm models (Pratten and Wilson 1999; Pratten et al. 2000; Mcbain et al. 2005; Rasiah et al. 2005; Ledder et al. 2006; Filoche et al. 2007; Sissons et al. 2007; Pham et al. 2009; Azevedo et al. 2011).
In agreement with previous authors, we feel that it will be important to reproduce experiments from any microcosm model using inoculums from multiple donors (Rasiah et al. 2005; Ledder et al. 2006). Our findings showed that microcosms were relatively stable within subjects over time, but there clearly were differences in species composition between subjects. Even though some species were common to all donors, the HOMIM system only looks at the 16S rDNA gene. There are likely to be many genetic differences between strains of the same species from different people, and those cannot be detected by HOMIM. Our ongoing studies therefore are being carried out with multiple donors, to more fully sample genotypic and phenotypic variation between microcosms.
This work was supported by NIH grant 1 R01 DE021366 from the National Institute for Dental and Craniofacial Research, Bethesda, MD, USA. Parts of this work were carried out in the University of Minnesota Characterization Facility, which receives partial support from the National Science Foundation. We thank Dr Bruce Paster, Sean Cotton and Alexis Kokaris of the Human Microbe Identification Microarray Service of the Forsyth Dental Institute, Boston, MA, for helpful comments regarding the interpretation of HOMIM data.