2.1

Preparation of artificial carious lesions by pH-cycling

Thirty extracted non-carious human third molars were obtained with patient informed consent under a protocol approved by the Human Assurance Committee of the Medical College of Georgia. A 1 mm thick dentin disk devoid of enamel and pulp exposure was prepared by making two parallel cuts perpendicular to the longitudinal axis of each tooth using a low-speed Isomet saw (Buehler, Lake Bluff, IL, USA) under water cooling. The surface for creating the partially-demineralized dentin was polished with 1200-grit silicon carbide paper under running water to create a smooth surface layer. The other surface of each dentin disk, together with the circumferential enamel rim and 1 mm of the peripheral dentin of the polished surface (to serve as mineral references) was painted with two coats of an acid-resistant varnish. A 300 ± 30 μm thick layer of partially-demineralized dentin was created on the uncoated surface by pH-cycling procedure to mimic caries-affected dentin below the excavated surface . After pH-cycling, each disk was sectioned transversely under running water to create two 3 mm wide slabs containing the partially-demineralized dentin.

2.2

Micro-CT scans

To assess the lesion depth and mineral loss, each slab was scanned non-destructively under water using a SkyScan 1174 compact X-ray micro-CT scanner (Micro Photonics, Allentown, PA, USA). A custom-made mold was created for each specimen using a sectioned pipette tip and polyvinylsiloxane impression material to serve as a positioning jig. The latter was attached perpendicularly to the specimen turntable of the micro-CT scanner to acquire the whole image of each slice. As the X-ray source is polychromatic, a 1 mm thick aluminum filter was placed in front of the detector to remove low-energy radiation. Scanning was performed with a spatial resolution of 6.28 μm at 50 kV and 800 μA. Flat-field correction, geometric correction for random movement during the acquisition phase and ring artifact correction during the reconstruction phase were performed to reduce ring artifacts. A 20% beam hardening correction was employed during the reconstruction phase using the NRecon software (Skyscan 1174).

Following image reconstruction, two-dimension virtual slices in the sagittal plane were acquired using the Data Viewer software (Skyscan 1174). The same acquisition and reconstruction parameters were used when the same slab was re-scanned during subsequent months. The sagittal virtual serial sections derived from each slab were subjected to maximum intensity projection (MIP) with the CTAnalyzer software (Skyscan1174) to obtain a stacked 2D image ( Fig. 1 ). As the CTAnalyzer only permits individual line scans across an image, the stacked 2D image was imported into ImageJ (NIH, Bethesda, MD, USA) to produce an overall mineral profile within a standardized volume of interest (VOI) ( Fig. 1 ). A white vertical line was formed extending from the radiopaque, non-demineralized part of the slab surface to the radiolucent lesion surface. This virtual line served as the superimposition reference for mineral profiles obtained during different time periods and eliminated the manifestation of streak artifacts introduced by incorporating a highly radiopaque reference within the specimen . Mineral profiles were determined at exactly the same area (4 × 1 mm) within the stacked image during the experiment ( Fig. 1 ). The length scale of the ordinate in the overall mineral profile was expressed in micrometers based on the resolution employed during image acquisition. The gray-scale attenuation values in the abscissa was expressing as the relative mineral volume by normalizing the mineral density of the mineralized dentin to 50 vol.% mineral density . The output parameters obtained were mineral content profiles of the lesions, lesion depth (μm, being the depth where the relative mineral content was 95% of the mineralized dentin value) and the integrated mineral loss (Δ Z , vol.% μm) as described previously .

A schematic illustrating how the overall mineral profile of an artificial carious lesion was derived from the volume of interest using stacked 2D micro-CT scanned images.

2.3

STMP pre-treatment and dentin bonding

Based on the baseline micro-CT data, 32 slabs with 300 ± 30 μm lesion depth were selected for the following experiments. Slabs with lesion depths that fell outside this range were rejected. As protein phosphorylation with STMP is usually performed at pH > 11 to open its closed ring structure , STMP (Mw 305.9, Sigma–Aldrich, St. Louis, MO, USA) was hydrolyzed at pH 12 for 5 h followed by neutralization to pH 7.4 before use. Based on the adsorption characteristics of STMP on demineralized dentin collagen , 16 slabs were treated with 2.5 wt.% STMP solution for 5 min and randomly divided into two groups ( N = 8): (1) without adhesive; (2) with adhesive. The other 16 slabs were not treated with hydrolyzed STMP and were designated as the negative controls for the no-adhesive with adhesive infiltrated experimental groups. For both the adhesive experimental group and its corresponding control group, five coats of an acetone-based unfilled adhesive (One-Step, Bisco Inc., Schaumburg, IL, USA) were liberally applied to the artificial carious lesion, allowed to soak in the dark for 1 min to permit as much adhesive infiltration as possible into the thick layer of partially-demineralized dentin and light-polymerized for 20 s. As the lesions were already partially demineralized, no additional phosphoric acid etchant was employed prior to the application of the two-step etch-and-rinse adhesive. The adhesive infiltrated slabs were stored at 100% relative humidity for 24 h to enable the bonds to mature. Slabs from the non-adhesive groups were stored in deionized water during that period.

2.4

Remineralization medium

A 1 cm diameter, 2 mm thick resin composite disk consisting of a light-polymerizable hydrophilic resin blend (50 wt.%), 45 wt.% set white Portland cement powder and 5 wt.% silanized silica was used as the sustained releasing sources of calcium and hydroxyl ions . The set white Portland cement was prepared by mixing with water in a 0.35–1 water–powder ratio, allowed to set completely for 1 week in 100% relative humidity, ground to a fine powder and sieved to retrieve particles that were smaller than 38 μm in diameter. A simulated body fluid (SBF) was prepared by dissolving 136.8 mM NaCl, 4.2 mM NaHCO ₃ , 3.0 mM KCl, 1.0 mM K ₂ HPO ₄ ·3H ₂ O, 1.5 mM MgCl ₂ ·6H ₂ O, 2.5 mM CaCl ₂ and 0.5 mM Na ₂ SO ₄ in deionized water and adding 3.08 mM sodium azide to prevent bacterial growth. The SBF was buffered to pH 7.4 with 0.1 M Tris Base and 0.1 M HCl and served as the control remineralization medium. For the biomimetic remineralization medium, 500 μg/mL polyacrylic acid (PAA, Mw 1800; Sigma–Aldrich) was added to the SBF and buffered to pH 7.4 to stabilize the amorphous calcium phosphate produced by the interaction of the set Portland cement with SBF in the form of nanoprecursor droplets .

2.5

Remineralization of artificial carious lesions

Each slab in the experimental and control groups was placed on top of a composite disk inside a glass scintillation vial with lesion surface contacting the composite. This simulated the placement of a composite over the adhesive infiltrated or unbonded lesion but alleviated the difficulty in removing the Portland cement-containing composite during subsequent ultramicrotomy. The assembly was placed inside a glass scintillation vial, which was filled with 15 mL of biomimetic remineralization medium for the two experimental groups or 15 mL of SBF for the two control groups. The specimens were incubated in each medium for four months at 37 °C, with the medium refreshed twice every month. The pH was monitored daily so that the value was above 9.5 to enable the formation of apatite instead of octacalcium phosphate from the initially formed amorphous calcium phosphate phase . Each slab was retrieved at monthly intervals (i.e. 1, 2, 3 and 4 months), rinsed briefly with water to remove loose surface precipitates, inserted into the corresponding micro-CT positioning jig, covered with pH-adjusted deionized water (pH 7.4) and scanned with the micro-CT. After scanning, each slab was removed from the positioning jig and returned to the corresponding immersion medium for continuous mineralization.

2.6

Statistical analyses

The 32 slabs with artificial carious lesions (mean lesion depth 300 ± 30 μm) were assigned to the two experimental groups and the two control groups ( N = 8) by analyzing the slab assignment with one-way ANOVA at α = 0.05 to ensure there were no differences in the baseline lesion depths among the four groups prior to the commencement of the experiment.

As there was no remineralization observed in the two control groups, the lesion depths, integrated mineral losses (Δ Z s), and the relative mineral contents along the lesion surface at baseline and after 4 months of remineralization in the non-adhesive and adhesive experimental groups were statistically analyzed. As the normality (Shapiro–Wilk test) and homoscedasticity (Levene test) assumptions of the data derived from each of the three parameters appeared to be violated, each parameter was analyzed separately using the Mann–Whitney rank sum test at α = 0.05.

2.7

Transmission electron microscopy (TEM)

After remineralization for four months, four specimens from each of the four experimental and control groups were selected for transmission electron microscopy. The selection criteria were that the overall lesion depths and changes in Δ Z were closest to the mean values obtained for that particular group. The slabs were fixed in Karnovsky’s fixative, post-fixed in 1% osmium tetroxide, dehydrated in an ascending ethanol series (50–100%), immersed in propylene oxide as a transitional fluid and embedded in epoxy resin.

Non-demineralized thick sections (210–250 nm thick) of the entire artificial carious lesion including part of the mineralized dentin base were first prepared for overall evaluation of the effect of remineralization. The blocks were then trimmed to 1.5 × 1.5 mm for preparation of non-demineralized thin sections (90–110 nm thick). These sections were prepared by diamond knives mounted on a Leica EM UC6 ultramicrotome (Leica Microsystems GmbH, Wetzlar, Germany), collected on single-slot carbon and formvar-coated nickel grids and examined without further staining. After initial examination of the thin sections, selected specimens of interest were demineralized in situ by placing the grid upside down over a drop of 0.05 N HCl for 60 s. After demineralization, each grid was dipped in-and-out of deionized water to remove precipitates and then stained with 1% phosphotungstic acid followed by 2% uranyl acetate to examine the status of collagen within the artificial carious lesions. In addition, two rejected slabs (lesion depths beyond the range required) were completely demineralized in formic acid/sodium formate within one week after preparation of the artificial carious lesions, embedded in epoxy resin as described previously and prepared for stained thin sections for examination of the status of the collagen within the artificial carious lesion prior to remineralization. Examination was performed using a JEM-1230 TEM (JEOL, Tokyo, Japan) at 110 kV.

2.1

Preparation of artificial carious lesions by pH-cycling

Thirty extracted non-carious human third molars were obtained with patient informed consent under a protocol approved by the Human Assurance Committee of the Medical College of Georgia. A 1 mm thick dentin disk devoid of enamel and pulp exposure was prepared by making two parallel cuts perpendicular to the longitudinal axis of each tooth using a low-speed Isomet saw (Buehler, Lake Bluff, IL, USA) under water cooling. The surface for creating the partially-demineralized dentin was polished with 1200-grit silicon carbide paper under running water to create a smooth surface layer. The other surface of each dentin disk, together with the circumferential enamel rim and 1 mm of the peripheral dentin of the polished surface (to serve as mineral references) was painted with two coats of an acid-resistant varnish. A 300 ± 30 μm thick layer of partially-demineralized dentin was created on the uncoated surface by pH-cycling procedure to mimic caries-affected dentin below the excavated surface . After pH-cycling, each disk was sectioned transversely under running water to create two 3 mm wide slabs containing the partially-demineralized dentin.

2.2

Micro-CT scans

To assess the lesion depth and mineral loss, each slab was scanned non-destructively under water using a SkyScan 1174 compact X-ray micro-CT scanner (Micro Photonics, Allentown, PA, USA). A custom-made mold was created for each specimen using a sectioned pipette tip and polyvinylsiloxane impression material to serve as a positioning jig. The latter was attached perpendicularly to the specimen turntable of the micro-CT scanner to acquire the whole image of each slice. As the X-ray source is polychromatic, a 1 mm thick aluminum filter was placed in front of the detector to remove low-energy radiation. Scanning was performed with a spatial resolution of 6.28 μm at 50 kV and 800 μA. Flat-field correction, geometric correction for random movement during the acquisition phase and ring artifact correction during the reconstruction phase were performed to reduce ring artifacts. A 20% beam hardening correction was employed during the reconstruction phase using the NRecon software (Skyscan 1174).

Following image reconstruction, two-dimension virtual slices in the sagittal plane were acquired using the Data Viewer software (Skyscan 1174). The same acquisition and reconstruction parameters were used when the same slab was re-scanned during subsequent months. The sagittal virtual serial sections derived from each slab were subjected to maximum intensity projection (MIP) with the CTAnalyzer software (Skyscan1174) to obtain a stacked 2D image ( Fig. 1 ). As the CTAnalyzer only permits individual line scans across an image, the stacked 2D image was imported into ImageJ (NIH, Bethesda, MD, USA) to produce an overall mineral profile within a standardized volume of interest (VOI) ( Fig. 1 ). A white vertical line was formed extending from the radiopaque, non-demineralized part of the slab surface to the radiolucent lesion surface. This virtual line served as the superimposition reference for mineral profiles obtained during different time periods and eliminated the manifestation of streak artifacts introduced by incorporating a highly radiopaque reference within the specimen . Mineral profiles were determined at exactly the same area (4 × 1 mm) within the stacked image during the experiment ( Fig. 1 ). The length scale of the ordinate in the overall mineral profile was expressed in micrometers based on the resolution employed during image acquisition. The gray-scale attenuation values in the abscissa was expressing as the relative mineral volume by normalizing the mineral density of the mineralized dentin to 50 vol.% mineral density . The output parameters obtained were mineral content profiles of the lesions, lesion depth (μm, being the depth where the relative mineral content was 95% of the mineralized dentin value) and the integrated mineral loss (Δ Z , vol.% μm) as described previously .

A schematic illustrating how the overall mineral profile of an artificial carious lesion was derived from the volume of interest using stacked 2D micro-CT scanned images.

2.3

STMP pre-treatment and dentin bonding

Based on the baseline micro-CT data, 32 slabs with 300 ± 30 μm lesion depth were selected for the following experiments. Slabs with lesion depths that fell outside this range were rejected. As protein phosphorylation with STMP is usually performed at pH > 11 to open its closed ring structure , STMP (Mw 305.9, Sigma–Aldrich, St. Louis, MO, USA) was hydrolyzed at pH 12 for 5 h followed by neutralization to pH 7.4 before use. Based on the adsorption characteristics of STMP on demineralized dentin collagen , 16 slabs were treated with 2.5 wt.% STMP solution for 5 min and randomly divided into two groups ( N = 8): (1) without adhesive; (2) with adhesive. The other 16 slabs were not treated with hydrolyzed STMP and were designated as the negative controls for the no-adhesive with adhesive infiltrated experimental groups. For both the adhesive experimental group and its corresponding control group, five coats of an acetone-based unfilled adhesive (One-Step, Bisco Inc., Schaumburg, IL, USA) were liberally applied to the artificial carious lesion, allowed to soak in the dark for 1 min to permit as much adhesive infiltration as possible into the thick layer of partially-demineralized dentin and light-polymerized for 20 s. As the lesions were already partially demineralized, no additional phosphoric acid etchant was employed prior to the application of the two-step etch-and-rinse adhesive. The adhesive infiltrated slabs were stored at 100% relative humidity for 24 h to enable the bonds to mature. Slabs from the non-adhesive groups were stored in deionized water during that period.

2.4

Remineralization medium

A 1 cm diameter, 2 mm thick resin composite disk consisting of a light-polymerizable hydrophilic resin blend (50 wt.%), 45 wt.% set white Portland cement powder and 5 wt.% silanized silica was used as the sustained releasing sources of calcium and hydroxyl ions . The set white Portland cement was prepared by mixing with water in a 0.35–1 water–powder ratio, allowed to set completely for 1 week in 100% relative humidity, ground to a fine powder and sieved to retrieve particles that were smaller than 38 μm in diameter. A simulated body fluid (SBF) was prepared by dissolving 136.8 mM NaCl, 4.2 mM NaHCO ₃ , 3.0 mM KCl, 1.0 mM K ₂ HPO ₄ ·3H ₂ O, 1.5 mM MgCl ₂ ·6H ₂ O, 2.5 mM CaCl ₂ and 0.5 mM Na ₂ SO ₄ in deionized water and adding 3.08 mM sodium azide to prevent bacterial growth. The SBF was buffered to pH 7.4 with 0.1 M Tris Base and 0.1 M HCl and served as the control remineralization medium. For the biomimetic remineralization medium, 500 μg/mL polyacrylic acid (PAA, Mw 1800; Sigma–Aldrich) was added to the SBF and buffered to pH 7.4 to stabilize the amorphous calcium phosphate produced by the interaction of the set Portland cement with SBF in the form of nanoprecursor droplets .

2.5

Remineralization of artificial carious lesions

Each slab in the experimental and control groups was placed on top of a composite disk inside a glass scintillation vial with lesion surface contacting the composite. This simulated the placement of a composite over the adhesive infiltrated or unbonded lesion but alleviated the difficulty in removing the Portland cement-containing composite during subsequent ultramicrotomy. The assembly was placed inside a glass scintillation vial, which was filled with 15 mL of biomimetic remineralization medium for the two experimental groups or 15 mL of SBF for the two control groups. The specimens were incubated in each medium for four months at 37 °C, with the medium refreshed twice every month. The pH was monitored daily so that the value was above 9.5 to enable the formation of apatite instead of octacalcium phosphate from the initially formed amorphous calcium phosphate phase . Each slab was retrieved at monthly intervals (i.e. 1, 2, 3 and 4 months), rinsed briefly with water to remove loose surface precipitates, inserted into the corresponding micro-CT positioning jig, covered with pH-adjusted deionized water (pH 7.4) and scanned with the micro-CT. After scanning, each slab was removed from the positioning jig and returned to the corresponding immersion medium for continuous mineralization.

2.6

Statistical analyses

The 32 slabs with artificial carious lesions (mean lesion depth 300 ± 30 μm) were assigned to the two experimental groups and the two control groups ( N = 8) by analyzing the slab assignment with one-way ANOVA at α = 0.05 to ensure there were no differences in the baseline lesion depths among the four groups prior to the commencement of the experiment.

As there was no remineralization observed in the two control groups, the lesion depths, integrated mineral losses (Δ Z s), and the relative mineral contents along the lesion surface at baseline and after 4 months of remineralization in the non-adhesive and adhesive experimental groups were statistically analyzed. As the normality (Shapiro–Wilk test) and homoscedasticity (Levene test) assumptions of the data derived from each of the three parameters appeared to be violated, each parameter was analyzed separately using the Mann–Whitney rank sum test at α = 0.05.

2.7

Transmission electron microscopy (TEM)

After remineralization for four months, four specimens from each of the four experimental and control groups were selected for transmission electron microscopy. The selection criteria were that the overall lesion depths and changes in Δ Z were closest to the mean values obtained for that particular group. The slabs were fixed in Karnovsky’s fixative, post-fixed in 1% osmium tetroxide, dehydrated in an ascending ethanol series (50–100%), immersed in propylene oxide as a transitional fluid and embedded in epoxy resin.

Non-demineralized thick sections (210–250 nm thick) of the entire artificial carious lesion including part of the mineralized dentin base were first prepared for overall evaluation of the effect of remineralization. The blocks were then trimmed to 1.5 × 1.5 mm for preparation of non-demineralized thin sections (90–110 nm thick). These sections were prepared by diamond knives mounted on a Leica EM UC6 ultramicrotome (Leica Microsystems GmbH, Wetzlar, Germany), collected on single-slot carbon and formvar-coated nickel grids and examined without further staining. After initial examination of the thin sections, selected specimens of interest were demineralized in situ by placing the grid upside down over a drop of 0.05 N HCl for 60 s. After demineralization, each grid was dipped in-and-out of deionized water to remove precipitates and then stained with 1% phosphotungstic acid followed by 2% uranyl acetate to examine the status of collagen within the artificial carious lesions. In addition, two rejected slabs (lesion depths beyond the range required) were completely demineralized in formic acid/sodium formate within one week after preparation of the artificial carious lesions, embedded in epoxy resin as described previously and prepared for stained thin sections for examination of the status of the collagen within the artificial carious lesion prior to remineralization. Examination was performed using a JEM-1230 TEM (JEOL, Tokyo, Japan) at 110 kV.