In vitro simulation of erosive challenges to human enamel using a novel artificial mouth

Abstract This in vitro work reports upon the design, build and operation of an artificial environment (Saltus) that sought to simulate the process of in vivo dental erosion upon human enamel. A novel testing environment, housed 8 erosion testing substrate specimens, that on separate occasions were subject to 4 different experimental diets, of increasing erosive challenge, simulating the consumption of an acidic beverage. Each set of specimens was subjected to one of the experimental diets only. These were liquid only and administered the test beverage over a standardized range of volumes and durations. Flow of both artificial unstimulated and stimulated saliva was maintained throughout and the effects upon the substrates were measured by profilometry, surface microhardness determination and chemical analysis of the saliva and beverage mixture for traces of Calcium and Phosphate ions. The overall trend of surface hardness reduction, depth of surface loss and ion loss across the diets increased in proportion to the severity of insult. Accepting the limitations of this study Saltus appeared to perform well as an environment in which to simulate and assess dental erosion using parameters defined by previous in vivo observations of human drinking behaviour. The authors however acknowledge that in vitro testing can never replicate fully the in vivo situation.

In any in vitro model, a guided standardization of experimental protocols can easily be made, which offers the potential to examine one variable at a time or, if desired, other variables can be introduced into the model. Furthermore, the ability to accurately measure an eroded tooth surface in vitro using precise devices and techniques favors such an approach compared to the in vivo setting.
In general, a lack of a standardization among different in vitro and in situ erosion testing experiments render them non comparable as a result of different experimental variables (West et al., 2011). In an endeavor to overcome this, parameters for a realistic erosion testing regime based upon natural human drinking behaviour have been developed derived from behavioral observations (Qutieshat, Mason, & Chadwick, 2015). Often in vitro erosion testing regimes use extracted human teeth as the substrate upon which the effects of dental erosion are observed. These often use unrealistically long exposure times to gain a measurable effect, that do not take account of normal drinking behaviour and effects of saliva (Cochrane, Cai, Yuan, & Reynolds, 2009;Jensdottir, Bardow, & Holbrook, 2005). In the past an artificial mouth has been used to assess the effect of mineral supplements to citric acid upon bovine enamel but the basis of its protocol of operation, that sought to mimic normal sipping behaviour, was not given (Attin, Meyer, Hellwig, Buchalla, & Lennon, 2003).
The purpose of this work was to report upon the design, build and operation of an environment that sought to simulate in vitro the process of in vivo dental erosion upon human dental enamel. It aimed to permit control of the variables of beverage consumption volume and flow rate as well as both the composition and flow rate of artificial saliva.   They were stored in distilled water following disinfection in sodium hypochlorite. The roots of each tooth were removed just above the cemento-enamel junction and the remaining coronal portion of the tooth was embedded in acrylic resin aligning it vertically in cylindrical molds 3 cm in diameter. Sagittal slices, 1 mm thick, were cut in a mesialdistal direction using a slow speed diamond saw (Isomet  (Qutieshat et al., 2015) for uniform flow) Temperature = 14.9°C For full dose dietstotal daily volume = 660 ml, Administration time of beverage = 44 minutes.
For half dose dietstotal daily volume = 330 ml, Administration time of beverage = 22 minutes. Stimulated saliva flow rate (Dawes, 1970;Dawes, 1987) 5.0 Unstimulated saliva flow rate in waking hours (Thomson, Chalmers, Spencer, & Slade, 2000) 0.3 Unstimulated saliva flow rate in sleeping hours (Thomson et al., 2000) 0.1 Buehler Ltd., USA), at 450 RPM, under water coolant until the first signs of enamel were observed in the sections. Thereafter 3 mm thick slices were prepared. Only samples containing no visible dentine were selected for use in this work following lapping the specimen flat using a PM5 precision lapping and polishing machine (Longitech, Glasgow, Scotland) and a slurry of calcined aluminium oxide powder with a particle size of 9 μm (Longitech, Glasgow, Scotland). Such samples were placed on the sample holders of Saltus. A typical sample of enamel acquired in this manner measured 7 × 7 mm at its exposed surface.

| Characterization of enamel specimens
Prior to and following exposure to an erosive beverage the surface hardness and surface profile of each specimen were determined.
Surface micro hardness -A TIV (Through Indenter Viewing) hardness tester (GE Measurement & Control, Groby, UK) with Vickers diamond, under a 9.8 N load, was used to measure the surface hardness at 10 sites per specimen. From these the percentage relative hardness of each specimen was calculated using the formula.

| Erosion testing regime
In this work a range of drinking behaviors, that exposed the erosion test substrate to an erosive challenge, were simulated and their effects upon human enamel were investigated. These looked at both different daily volumes of consumption and their duration. The erosion cycles, termed diets, used were informed by a previously published observational study (Qutieshat et al., 2015) and were delivered to the specimens using Saltus. For the purposes of this work a diet is a program delivered by Saltus that is comprised of a series of cycles through which erosion substrates were exposed to artificial saliva and a test beverage. Diets were of the duration of 5 or 7 days. For each diet the experiment was repeated twice upon new specimens. Operation of Saltus was continuous over each 24 hour period of operation and consisted of 3 daily periods; • A day "waking hours" period It should be noted that for all of these diets a rest cycle comprised a 24 hour period during which unstimulated saliva flowed at a rate of 0.3 ml min −1 , between the hours of 0700 to 2300 hours (Waking Hours), and from 2300 until 0700 hours (Sleeping Hours) at 0.1 mm min −1 . These primarily served to facilitate ion collection rather than imitate human behaviour. On days when a test beverage was introduced into the system (Test Cycle) this pattern continued with the exception of between the hours of 1400-1500 during which stimulated saliva flowed at 5.0 ml min −1 and the test beverage was introduced according to the diet being administered.
To investigate any effects of diets (diet 1, 2, 3 and 4) and experimental run (Run 1 and 2) on the surface hardness change a 2-way analysis of variance on this data was undertaken. A follow-up analysis (Bonferroni post-test) was undertaken to determine the effect of the diet factor over all the groups (diet 1, 2, 3 and 4).
To investigate any effects of diets (diet 1, 2, 3 and 4) and experimental run on the amount of surface loss a 2-way analysis of variance on this data was undertaken. A follow-up analysis (Bonferroni posttest) was undertaken to determine the effect of the diet factor over all the groups (diet 1, 2, 3 and 4).
To investigate any effects of diets (diet 1, 2, 3 and 4) and ion type upon ion loss a 2-way analysis of variance on this data was undertaken.
A follow-up analysis (Bonferroni post-test) was undertaken to determine the effect of the diet factor over all the groups (diet 1, 2, 3 and 4).
For the purposes of comparison the observed depth of surface loss was divided by the duration of experimental run.

| RESULTS
Prior to the introduction of any erosion substrates into the device analysis of the Calcium and Phosphate ion concentrations of the collected saliva mixture following a substrate and beverage free one day diet confirmed no effect upon the ionic composition of the saliva.
The % absorption spectrophotometer readings for bromophenol dye mixed with 1 M citric acid, using either the magnetic stirrer or Saltus, did not differ significantly (P > 0.05). This, together with the observation that the solution mixed in Saltus was yellow in color with no blue spotting, confirmed the efficiency of the devices mixing system.   Table 3 presents this data as a mean value of depth of surface loss per hour for each diet. It is clear that a dose response relationship exists between the quantity of consumption and the rate of surface loss. Figure 4 gives the overall mean ion loss (mmol/day) and their associated standard deviations for Calcium and Phosphate following the 4 diets.
For each measured parameter separate two way analyses of variance were undertaken to determine if the different diets, experimental runs and their interaction had any statistically significant effects. In all cases no effect overall of experimental run or interaction of diet and experimental run was found (P > 0.05). The effect of the different

FIGURE 2
The mean post diet hardness values and their associated standard deviations, for human enamel samples, for each of the 4 diets administered on two experimental runs as expressed as a % of the prediet hardness values. Note 100% represents no change in hardness value

FIGURE 3
The mean surface loss (μm) and their associated standard deviations, relative to pre-diet values, for human enamel samples, for each of the 4 diets administered on two separate experimental runs. Note: a positive loss value represents a loss of material

FIGURE 4
The overall mean ion loss (mmol/day) and their associated standard deviations for Calcium and Phosphate following the 4 diets diets was however highly statistically significant (P < 0.001) for all parameters. Bonferroni post testing revealed that: • The % surface hardness reductions differed significantly (P < 0.05) for diet 1 relative to diets 3 and 4 on both experimental runs. No other significant differences were found. There was thus generally good agreement between the experimental runs.
• The observed differences in surface loss was only statistically significant (P < 0.01) in the cases of Diet 1 versus Diets 3 & 4, Diet 2 versus Diet 3, Diet 3 versus Diet 4 following run 1. In the case of run 2 this was also the case with the exception of Diet 3 versus Diet 4 where there was no statistically significant difference. Reproducibility between the runs was therefore good.
• For all ions (Calcium and Phosphate), with the exceptions of Diet 1 versus Diet 2 and Diet 3 versus Diet 4 significant differences (P < 0.05) were seen between the diets.

| DISCUSSION
The primary aim of this study was to design, build and evaluate a novel artificial mouth for use in the evaluation of dental erosion in vitro based upon behavioral observations. It was therefore not the aim at this stage to use it to evaluate a range of drinking habits more to illustrate the models utility. Before considering the findings revealed by The work sought to reproduce observed behaviour (Qutieshat et al., 2015) and inevitably due to engineering constraints (matching pump speed and tubing diameters) the observed physiological values for beverage flow rate were adjusted to a deliverable value. This may be of interest to other researchers planning such devices.
Although the present work was run in a highly controlled fashion it can be argued that the extent of the erosive lesion may be affected by inter-structural variations of the specimens. This could arise for example by the presence of biological variation (e.g. different specimen sources, location and environmental history). In this study however, this is unlikely to be a factor for variability among specimens in terms of their hardness values standard deviations did not exceed 13.2% (Table 3). It has been previously reported in the literature that hardness deviations of up to 16% are to be expected for human enamel specimens (Turrsi, Messias, Corona, & Serra, 2010).
When surface loss is taken into account (Figure 3 and  (Cochrane et al., 2009;Jensdottir et al., 2005) adopted a prolonged exposure-by-immersion time period of 24 hours which, although registering an effect, was not founded upon normal physiology and drinking behaviour. Another study reported that calcium ion loss per hour of acidic exposure was 0.46 mmol/l (Larsen & Richards, 2002) and its findings were in accord with another study (Xavier, Rai, Hegde, & Shetty, 2015) that utilized demineralisationremineralisation cycling and also demonstrated calcium ion loss of 0.43 mmol/l per hour. The same study also reported phosphate ion loss to be 0.52 mmol/l per hour. In the present work, human enamel calcium loss in Diets 1 and 2 was in the range of 18.2-20.7 mmol/l and the phosphate loss was in the range of 25.6-30.9 mmol/l. This is in line with the conclusion of an in vitro study that assessed human enamel mineral loss upon the exposure to a regular cola beverage.
Here the dental substrates were found to lose a stable ratio of calcium and phosphate throughout the erosive process (Willershausen & Schulz-Dobrick, 2004). In the present work a dose-response relationship was observed where single-dosed diets lost significantly less ions relative to double-dosed diets.
Accepting the limitations of this study Saltus performed well as an environment in which to simulate and assess pure dental erosion though it is acknowledged that a contemporary definition of erosive tooth wear (Lussi & Carvalho, 2014) describes it as a chemical mechanical phenomenon and the latter is absent in the model. Unlike a previously reported artificial mouth (Attin et al., 2003) used to assess the effect of mineral supplements to citric acid upon bovine enamel erosion, it permitted the simultaneous flow of different types of artificial saliva (stimulated and unstimulated) and erosive beverage and as such, together with the evidence based drinking behaviour that informed its operation (Qutieshat et al., 2015), offers an improvement in such laboratory simulation. Notwithstanding this it is hoped that the techniques described in its design will be of value to other researchers.
The authors however acknowledge that in vitro testing can never fully replicate the in vivo situation but this apparatus offers an improvement in such simulation.