Qualitative and quantitative differences in the subgingival microbiome of the restored and unrestored teeth

Background and Objective Metal‐based dental restorations with a subgingival outline may enhance plaque accumulation and bacterial colonization. This study aimed to investigate whether metal‐based restorations influence the composition of subgingival microbiome. Material and Methods Per subject one site with a metal‐based restoration and one contra‐lateral site without a restoration were selected on basis of radiographic bone loss ≤2 mm, restoration outline at sulcus level/subgingivally, pocket depth ≤4 mm, and no root canal treatments. Subgingival samples were collected with sterile paper‐points, and microbial profiles were obtained by 16S rRNA gene amplicon sequencing. Restorations were sampled with an Arkansas‐stone and the metal composition was determined using energy‐dispersive X‐ray spectroscopy. Results A total of 22 sites from 11 subjects were included. No significant differences for the clinical parameters were found between the restored and unrestored sites. The average age of the restorations was 14.9 ± 7.1 years. Firmicutes was the most prevalent phylum at the restored sites (32% vs 20% of the reads of the unrestored sites, P = 0.016), and Actinobacteria at the unrestored sites (33% vs 18% of the reads of the restored sites, P = 0.01). Overall, sequences clustered into 573 operational taxonomic units (OTUs). Species richness of the restored sites was significantly higher than species richness of the unrestored sites (117 ± 32 and 96 ± 20 OTUs, respectively, P = 0.013). No associations between the metal composition and bacterial profiles were found. Conclusion This study shows that metal‐based restorations may enhance colonization of Firmicutes and the neighboring pocket may harbor more diverse microbial communities.


| INTRODUC TI ON
The placement of a metal-based dental restoration with a subgingival outline may enhance plaque accumulation and bacterial colonization because of rough surface of the restorative material 1 and overhang of the restoration. 2 Gingival tissue cells respond to the bacteria by an inflammatory reaction, which in susceptible subjects may lead to breakdown of periodontal tissues. 3 It has also been shown that overhanging restorations play a role in the extent of periodontal breakdown. 2,4,5 Overhanging restorations result in anatomical changes, which may create a more favorable environment for periodontitis-associated bacteria. 6,7 Furthermore, galvanic corrosion due to a lower of pH in the crevice 8 may result in leakage of metal ions into the gingival crevice and tissues. 9,10 As a result, this could influence bacterial adhesion, toxicity and induce allergies. 11 On the other hand, metals such as silver (Ag), copper (Cu), gold (Au), and zinc (Zn) have antimicrobial properties and are used for centuries in dentistry and general health care. [12][13][14] Gingivitis is the first sign of inflammation of periodontal tissues.
About 50%-100% of the adult population can be characterized as having gingivitis. [15][16][17] Clinical signs of gingivitis are bleeding on probing (BoP), redness and swelling, and gingivitis occurs as a reaction of the gingival tissues to accumulation of bacteria and their toxic products on teeth, which are non-shedding surfaces. During the development of gingivitis, a shift in the microbiological composition can be seen; gram-positive cocci and rods are being replaced by gramnegative cocci, rods, filaments, fusobacteria, and spirochetes. 18,19 To date, our understanding of the microbiology of gingivitis and microbiological changes in relation to dental restorations is based on traditional targeted techniques. These techniques study the microbiota in a specific manner, that is, searching for specific known microorganism only. In vivo studies, using targeted techniques, have shown that Streptococcus and Actinomyces are the first colonizers of the tooth enamel surface. 20 With the introduction of openended techniques, it is possible to analyze thousands of bacterial sequences per sample. This technique provides an opportunity to study the entire composition of bacterial communities and identify potentially novel bacterial species in the subgingival biofilm. 18,21,22 On the basis of 16S rRNA, Griffen et al 21 identified several new taxa in subgingival plaque of periodontitis patients and patients with a healthy periodontium. Further, they reported that a distinction could be made between subgingival microbiomes of periodontally healthy and diseased sites.
The metal-based restorations and their actual metal composition may lead to alterations in the subgingival microbiome. We hypothesized that a subgingivally placed metal-based restoration may influence the composition of the subgingival microbiome. Therefore, the main aim of the present study was to investigate whether and how the subgingival microbiome differs at the sites with a metal-based restoration compared to the unrestored sites. Furthermore, association of other local factors such as BoP, extent of overhang, age of the restoration, and metal composition, as well as the systemic factor smoking with the subgingival microbiome was analyzed.

| Subject selection
The current study was designed as a cross-sectional split-mouth cohort study and included subjects who visited the clinics of the Academic Centre for Dentistry Amsterdam (ACTA). On the basis of <2 years old existing radiographs (peri-apical, bitewings or orthopantomograms), subjects were screened for alveolar bone loss ≤2 mm at the targeted sites. The following variables were extracted from the subjects electronic health record: (a) age and gender, (b) diabetes mellitus, (c) smoking habits, and (d) any known (metal) allergies. A subject was defined as a smoker if a current smoker or stopped smoking <1 year ago and as a non-smoker if a never smoker or stopped smoking ≥1 year ago. Subjects were excluded if they had taken antibiotics or received periodontal treatment in the last 6 months. The Medical Ethical Committee of the VU Medical Centre (VUMC), Amsterdam, approved the study (11/306). All subjects gave a written informed consent, and the study was carried out in accordance with the Declaration of Helsinki.

| Selection of the test and control sites
Non-molar and flat molar surfaces were selected when (a) one tooth site was restored with a metal-based material (porcelain-fused-tometal (PFM) crown or amalgam restoration), (b) the outline of the restoration was located at sulcus level or subgingivally, (c) pocket depth was not >4 mm, and (d) no root canal treatment was present. First, the restored site was selected as a test site (eg, tooth 12 mesio-buccal site) and then the unrestored site of the contra-lateral tooth (eg, tooth 22 mesio-buccal site) was selected as a control site.
Incisors were the first choice, and then the teeth were checked to the posterior direction until a tooth with a metal restoration was found. If the same site of the contra-lateral tooth was also restored, the next tooth was selected if unrestored (control site). No metalbased restorations in the tooth adjacent to the unrestored site were allowed.

| Microbial and metal sample collection
Cotton rolls were placed next to the sample sites, and the sampling area was gently air-dried. Supragingival plaque was carefully removed with a microbrush (Microbrush international, Grafton, WI, USA). The microbiological sample of subgingival plaque was taken from the test and control sites using two sterile paper-points per site (PP; Absorbent Points # 5-4; Henry Schein U.K. Holdings Ltd., Southall, Middlesex, UB2 4AU England). The paper-points were inserted into the pocket for 10 seconds, placed in an empty Eppendorf tube and stored at −80°C until DNA extraction.
After microbial sampling and clinical measurements, a sample of the metal restoration was taken with a new Arkansas-stone (Dura-White FL2, Shofu, Japan) at 1000 rpm without water cooling. The restoration sample spot was polished and cleaned after sampling.

| Clinical measurements
The clinical measurements at test and control sites were performed after microbiological sampling except for plaque which was recorded before removal of supragingival plaque. Plaque was scored as follows: 0, no plaque; 1, plaque only detectable with a probe; and 2, plaque visually detectable (modified from, 23 mPI). BoP was scored as follows: 0, no bleeding after probing; 1, bleeding after probing. Mobility was scored according to the Miller classification. 24 Furthermore, probing pocket depth (PPD), amount of gingival recession (REC), clinical attachment level (CAL), and suppuration were determined at sampled sites. At the test site, the amount of overhang was determined and categorized as follows: −1 = under-filled, 0 = no overhang, 1 = overhang only diagnosed with the use of a probe, 2 = overhang clinically visible, and 3 = overhang visible on the radiograph. The distance from the most apical part of the restoration outline until the gingival margin was measured. Also the age of the restoration and the type of restoration were assessed. All clinical examinations were performed by the same trained clinician (SR).

| DNA extraction, amplicon preparation, and pyrosequencing
DNA from subgingival plaque was extracted as described previously 25 and quantified on 16S rDNA content through real-time PCR. 26 Samples with negative PCR results were subjected to concentration of the DNA through vacuum centrifugation (Martin Christ Gefriertrocknungsanlagen GmbH, Osterode am Harz, Germany). All clinical samples were adjusted to the end DNA concentration of 20 pg/ μL and stored at −20°C until further analysis. Besides the clinical samples, DNA extracts from duplicate sterile paper-points and extraction blanks were included to control for a potential contamination. 27 Barcoded amplicon libraries of the small subunit ribosomal RNA (16S rRNA) gene hypervariable region V5-V7 were generated for each individual sample, pooled and sequenced by means of the Genome Sequencer FLX Titanium system (Roche, Basel, Switzerland) as described previously. 26 The pyrosequencing data were processed and quality filtered; reads containing ambiguous base calls, >1 error in the forward primer, >2 error in the reverse primer, >1 error in the barcode, >6 nt homopolymer sequence, the average quality score below 30, or a length <200 bp or >1000 bp were removed from the analyses. Sliding window test (50 nt) of quality scores was enabled, and sequences of low quality were truncated at the beginning of the poor quality window. 26 The cleaned reads were clustered in operational taxonomic units (OTUs) at a minimal sequence similarity of 97%, and the representative sequence of each cluster was assigned a taxonomy as described previously. 26

| Analyses of the metals
A modified method for the determination of the composition of the alloys in dental restorations was used. 28,29 In brief, the composition of the metal particles on the Arkansas-stone was determined both qualitatively and quantitatively by Energy-dispersive X-ray Spectroscopy (EDS) in combination with a high vacuum Scanning Electron Microscopy (SEM) (XL20, Philips/FEI, Eindhoven, The Netherlands). The metal particles of the Arkansas-stone were transferred onto a carbon tape. Energy of the electron beam was set at 30 keV with a spot size of 6.0. In a high vacuum environment, electrons were fired upon the metal particles resulting in free electrons dispersing energy. The amount of free energy was analyzed with a backscatter electron detector.

| RE SULTS
After extensive screening (N = 558) of the Electronic Records of the dental school, a total of 22 subjects were included to the study. From these subjects, both 22 restored and 22 unrestored sites were sampled. A total of 3 subjects were excluded because of negative PCR yield of both restored and unrestored samples ( Figure S1). From the remaining 19 subjects, for 11 subjects both the restored and the unrestored site sample (paired) and for 8 subjects only one of the two samples passed the pyrosequencing quality control ( Figure S1). The mean age of the 19 subjects was 52 years (±SD 10.5 years; 9 males and 10 females), 13 were Caucasians, 6 were smokers, 2 had diabetes, and none of the subjects reported allergy to metals.

| Characteristics of the restored and unrestored sites
Clinical characteristics of the restored and unrestored sites did not differ significantly (Table 1). Mean mPI of the restored sites was 0.8 ± 0.6 compared to 0.9 ± 0.6 at the unrestored sites, and mean BoP on the restored 0.8 ± 0.7 compared to 0.7 ± 0.6 at the unrestored sites. The mean PPD at the restored sites was 2.8 ± 0.9 mm compared to 2.5 ± 0.9 mm at the unrestored sites. Also CAL and mobility scores were similar between the two types of sites. The average age of the restorations was 14.9 ± 7.1 years (range between 7 and 30 years). Table 2 gives an overview of the metal content of the restorations. From the 19 restorations, 14 were PFM crowns and 5 were amalgam restorations. The majority of the PFM crowns (57%) contained copper, 43% gold and silver, 36% zinc and 29% platinum, palladium and chrome. All of the five amalgam restorations contained copper, silver, and mercury, and 80% contained also tin.

| Sequencing output
Of the overall reads, 83% passed the quality control and the de-

| Sequencing results of unrestored vs restored paired samples
First, we addressed the primary aim of the study and compared the microbiomes of paired subgingival samples from 11 individuals with unrestored and restored sites. Actinobacteria was the most prevalent phylum at the unrestored sites (33% vs 18% of the reads at the restored sites, P = 0.01). Firmicutes was the most prevalent phylum at the restored sites (32% vs 20% of the reads at the unrestored sites, P = 0.016) (Figure 1).
At the genus level (Figure 2), Streptococcus was at a significantly higher proportion present at the restored sites (20% of the reads) compared to the unrestored sites (12% of the reads, P = 0.033). At the unrestored sites, the proportion of genus Actinomyces (10% of the reads) and family Propionibacteriaceae (2% of the reads) was significantly higher than at the restored sites of the same individuals (5% and 0.5% at the restored sites, respectively; P = 0.026 for both genera).
In total, the sequences could be clustered into 573 OTUs.
Significant difference was observed for the total number of OTUs (observed species richness) between the restored and unrestored sites: the restored sites harbored a mean number of 117 ± 32 OTUs in comparison to the 96 ± 20 OTUs at the unrestored sites (P = 0.013). Also the Chao-1 index that estimates the overall species richness was significantly higher at the restored sites (144 ± 41) than at the unrestored sites (115 ± 32) (P = 0.008). However, differences in the Shannon Diversity index did not reach statistical significance (P = 0.458) with 2.9 ± 0.81 at the restored and 2.7 ± 0.49 at the unrestored sites. OTUs of Fusobacterium, Actinomyces, and

| Sequencing results of all restored samples
In an additional analyses, microbial samples from all restored sites (N = 15, including 11 samples from the subjects with paired samples and 4 samples from the subjects with only a sample from a restored site, Figure S3) were analyzed.
At the restored sites, there was some spatial discrimination in PCoA between the microbial profiles with and without bleeding ( Figure S4 Figure 3B) and Enterococcus at the restored nonbleeding sites (P = 0.005).

| Sequencing results of all unrestored samples
An additional analyses were also done for all microbial samples from the unrestored sites (N = 15, including 11 samples from the subjects with paired samples and 4 samples from the subjects with F I G U R E 1 Relative abundance of the major bacterial phyla of the unrestored and restored sites (11 subjects with paired samples). The vertical line separates the individual unrestored sites on the left from the individuals restored sites on the right. The phylum Actinobacteria was significantly more abundant at the unrestored sites (P = 0.01) while the phylum Firmicutes was significantly higher at the restored sites (P = 0.016) F I G U R E 2 A bar chart of the mean relative abundance for the major bacterial genera (11 subjects with paired samples). The genus Streptococcus was significantly increased at restored sites (P = 0.033) while the genus Actinomyces and family Propionibacteriacea were significantly increased at unrestored sites (P = 0.026 for both genera). Error bars indicate standard deviation * a significant difference

| D ISCUSS I ON
The present study is the first study investigating the effect of metal- is observed. 31,32 In the present study, a typical subgingival microbiome of a shallow periodontal pocket was found. 33 However, there were differences between the restored and unrestored peri- gram-negative and anaerobic organisms spreading from overhanging fillings to non-overhanging fillings.
Metal-based dental restorations consist of metals like gold, silver, zinc, tin, copper, platinum, mercury, palladium, cobalt, nickel, and chromium. 36 These metals can be released and found in the oral tissues 9 and may influence the host immune response. 37 Next to affecting the immune system, several metals including mercury, 38 copper and nickel-chromium, 39  The oral cavity has a large microbial diversity. A previous study on bacterial diversity of 10 periodontally healthy individuals reported on 128 OTUs in a pooled subgingival plaque sample. 43 In our study, on average 117 and 96 OTUs were found in the subgingival plaque at the restored and unrestored sites, respectively, and the restored sites showed significantly increased number of species richness in comparison to the unrestored sites. Several studies have shown that periodontitis-associated bacterial communities have increased species richness (=higher taxonomic diversity). 22,44 The higher bacterial species richness of the restored sites may indicate that these sites are at a higher risk for further periodontal breakdown and require regular periodontal follow-up. Nevertheless, a large variability was found in the subgingival microbial community structures between and within the subjects. To further explore subgingival microbiomes in various conditions, the future studies should include a larger number of subjects.
In conclusion, the present study showed that metal-based restorations are associated with enhanced colonization of the bacterial phylum Firmicutes, and the bleeding restored sites with enhanced colonization of the phyla Bacteroidetes and Spirochetes.
Further, the neighboring pocket of the restored teeth may harbor more diverse microbial communities. We speculate that altered surface structure and roughness, enhanced retention, galvanic corrosion and leakage of metals may influence microbial composition.

ACK N OWLED G EM ENTS
We would like to thank Ing. Arie Werner from Department of Dental