Background

Peri-implant health is defined by the absence of erythema, redness, bleeding on probing, swelling and suppuration around dental implants. Despite high implant survival rates, different biological complications can affect osseointegrated implants, including peri-implant mucositis and peri-implantitis [58]. Peri-implantitis is an oral inflammatory process that affects the surrounding tissues of osseointegrated implants and results in the loss of supporting bone and destruction of soft tissues [99]. According to Daubert et al. [29] and Konstantinidis et al. [59], the prevalence of peri-implant disease range between 13 and 25%, leading to a significant increase of patient morbidity, economic burden and eventual implant loss [5, 32]. As is the case for periodontal diseases [67], the primary etiological factor in the development of peri-implant diseases is the biofilm, which is a complex synthropic microbial community consisting of adherent cells embedded within a matrix composed of extracellular polymeric substances [46]. However, some discrepancies have been identified in the biofilms from peri-implantitis and periodontitis sites [8, 79].

Approximately 700 species of Prokaryota have been identified in the oral cavity, predominantly ascribed to 12 phyla: Firmicutes, Fusobacteria, Proteobacteria, Actinobacteria, Bacteroidetes, Chlamydiae, Chloroflexi, Spirochaetes, SR1, Synergistetes, Saccharibacteria (TM7) and Gracilibacteria [93]. Specifically, the mouth contains distinct niches with dynamic microbial communities, including saliva, gingiva, the hard and soft tissues of teeth, the tonsil, the gingival sulcus, the throat, the hard and soft palates and the buccal mucosa [31], and eventually dental implants [23]. These surfaces—whose differing chemistry, topography and stability provide different habitats for microorganisms—are colonized preferentially by different bacteria via surface-attachments, movements and complex interactions, resulting in spatial compositional variability [72]. Furthermore, oral bacterial communities exist in multiplex dynamic equilibrium states, with large and rapid changes in composition and activity in a temporal dimension in response to environmental conditions  [44]. The microbial communities are in symbiosis with the host shaped by co-evolution, contributing to digestion and homeostasis, neurological signaling, regulating immune and endocrine functions, modifying metabolism and eliminating toxins  [37, 41]. Nevertheless, under certain conditions, several pathogenic strains that are usually dominated by commensal bacteria can proliferate [84], and even some commensal bacteria can transit to a pathogenic lifestyle via complex changes involving gene expression patterns, the core genome and the pan-genome [70, 90, 115].

Several studies aiming to compare oral bacterial communities under health and disease conditions have been conducted [14, 38, 56], some of them focused on characterizing peri-implant and periodontal microbiota and its pathological changes [36]. The bacterial profile associated with peri-implant disease was reviewed by Pérez-Chaparro et al. [85], Rakic et al. [94], Sahrmann et al. [98], Butera et al. [15], Gazil et al. [43] and Rodríguez-Archilla and Palma-Casiano [96]. These authors revealed that the core peri-implantitis microbiome is enriched in periodontal-inflammation related taxa, including Fusobacterium nucleatum, Parvimonas micra and Aggregatibacter actinomycetemcomitans, Prevotella intermedia, Prevotella nigrescens, Treponema denticola, Tannerella forsythia, Campylobacter rectus and Porphyromonas gingivalis. According to most published research findings, oral bacterial communities from healthy implants, peri-implantitis and periodontitis sites show contrasting diversity and composition [28, 43, 105, 119]. However, conclusions regarding the role of individual taxa in oral pathogenesis are difficult to draw due to the presence of confounding factors, variability in experimental designs, interindividual variability and complex ecological communities. Furthermore, bacterial actions are often secondary to immunologic imbalance [4, 75].

Due to the irreversible nature of peri-implantitis, prognosis-based early therapeutic intervention is the best strategy to arrest the progression of the disease and prevent implant failure [114]. Thus, a wide range of early diagnosis methods have been suggested [19, 22, 95]. In this sense, molecular techniques—particularly metagenomic Next Generation Sequencing tests (mNGS)—are powerful approaches for the characterization of microbial communities in the oral cavity, and also for the detection and surveillance of obligate or facultative oral pathogens associated with peri-implantitis [50, 103]. Considering the high variability in microbiota composition between different oral microbial habitats at multiple scales, the selection of appropriate sampling methods is a key aspect of implant disease risk assessment [6]. Sample collection via direct removal of transepithelial abutment and bacterial DNA isolation from transepithelial abutment surface could be considered the most representative sampling strategy, but this is a destructive method and the replacement of the transepithelial abutment with another is required (thus making sampling harder, increasing the economic burden and preventing the possibility of long-term monitoring of the bacterial community in that particular transepithelial abutment).

Crevicular fluid collection via sterile paper points (either from an adjacent tooth or from the studied transepithelial abutment) was initially regarded as a promising approach, since it is nondestructive, fast and easy to carry out, and allows long-term sampling at different time points. The question is, could these alternative sampling strategies be considered representative of bacterial communities formed on transepithelial abutment surfaces? Are microbial communities of samples collected using these methods significantly different than those obtained by sampling directly transepithelial abutments? In what respect and to what level? Thus, the aim of this preliminary research was to select the most appropriate sampling approach to characterize bacterial biofilms from transepithelial abutment surfaces. For that, three sampling strategies were compared: crevicular fluid from an adjacent tooth (ToCF), crevicular fluid from transepithelial abutment (TACF) and transepithelial abutment (TA), with a particular focus on peri-implant or periodontal health and disease related taxa.

Methods

This research is part of a randomized clinical trial that was designed to assess the effect of transepithelial abutment surface on the biofilm formation. The trial was registered at Trial Registration ClinicalTrials.gov under the number NCT03554876. The study protocol and informed consent, in full accordance with the ethical principles of the Declaration of Helsinki of 1975, as revisited in 2000, were approved by the ethical committee of investigation with medicines of the Basque Country (FIBEA-06-EC/17/Multi-Im). This study constitutes a partial analysis of the results of this clinical trial aiming to assess the effect of sampling strategies on metagenomic outcomes. Additional information regarding the experimental design is provided as supplementary data (Supplementary Material 1). For the analysis of the effect of sampling strategy, transepithelial abutments with a machined surface were included. Patient selection for this randomized controlled clinical trial was based on the following criteria:

  • Inclusion criteria

    • Patients with an age ≥ 18 years.

    • Need for the placement of at least 3 dental implants.

    • Complete mouth plaque index ≤ 20% and absence of active periodontal disease.

    • Bleeding index ≤ 30%.

    • Pocket probing depth at the adjacent teeth < 4 mm.

    • No use of antibiotics in the last 6 months.

    • Nonsmoker.

    • Possibility of attending all the planned visits.

    • Signing of informed consent.

  • Exclusion criteria

    • Has severe hematological disease.

    • Has received or receiving in the last 30 days at least one of the following treatments: radiotherapy, chemotherapy, Immunosuppressive therapies, systemic corticoids, and anticoagulants.

    • Presence of malignancy, hemangioma or angioma at the site where dental implants will be placed.

    • Patients receiving bisphosphonates (oral or systemic).

    • Presence of metabolic osseous disease.

    • Presence of diseases that affect the oral mucosa.

    • Presence of diabetes mellites.

    • Severe parafunctional habits and/or temporomandibular joint disorders.

    • Pregnancy or breast-feeding.

    • Physical or mental disability to maintain good oral hygiene.

    • Participating in other study.

    • Other disabilities to participate in the study.

Clinical procedure

A total of 12 patients received professional oral hygiene and instructed how to maintain a good oral hygiene. After implant insertion, transepithelial abutments with machined surface were connected to the implants. Two months later, microbiome samples were acquired using different strategies. Sterile paper points size 30 (Maillefer, Ballaigues, Switzerland) were utilized to collect periodontal crevicular fluid from adjacent teeth (ToCF) and peri-implant crevicular fluid (TACF). These samples were collected from at least one adjacent healthy tooth and also from each previously connected transepithelial abutment with machined surface. After crevicular fluid collection, transepithelial abutments (TA) were removed for processing. The collected samples were stored at -80 °C until bacterial DNA isolation.

DNA extraction

Total microbial metagenomic DNA from each sample was extracted using the DNeasy PowerBiofilm DNA isolation kit (Qiagen, Germany) following the manufacturer’s instructions. The strips were homogenized (1 cycle at 6400 rpm for 30 s) with a Precellys 24 Tissue Homogenizer (Bertin Technologies, France) and the implants were homogenized with an IKA MS 3 digital vortex (IKA, Germany) for 10 min at 2250 rpm. DNA quantification and quality control was performed using a Nanodrop 8000 (Thermo Fisher Scientific, MA, USA) and a Qubit fluorometer (Thermo Fisher Scientific, MA, USA). Extracted DNA was kept frozen at − 30 °C until library preparation.

Library preparation and sequencing

16S rRNA library preparation workflow for MiSeq sequencing platform was performed as suggested by Illumina. The specific primers for the 16S rRNA gene (v3- v4 region) were selected from Klindworth et al. [57] and combined with Illumina adapter overhang nucleotide sequences to obtain a single amplicon of approximately ~ 460 bp. After PCR product purification, dual index barcodes and Illumina sequencing adapters were attached using the Nextera XT v2 index kit (Illumina, CA, USA) as a previous step to pooling, libraries were quantified by LabChip GX touch HT nucleic acid analyzer together with DNA 5 K/RNA/CZE chip (PerkinElmer, MA, USA) and diluted for an estimated sequencing depth of ~ 100.000 reads per sample. finally, pooled libraries were denatured with NaOH and diluted with hybridization buffer (library loading concentration = 6 pm) before MiSeq (llumina, CA, USA) sequencing. PhiX was included in each run to serve as an internal control for these low diversity libraries. Paired-end sequencing (2 × 300 bp) was performed using MiSeq v3 reagent kits (600 cycles) (Illumina, CA, USA).

Data processing

Secondary analysis was performed on BaseSpace using the 16S metagenomics application (Illumina, CA, USA). After assembling, full-length sequences from paired ends were referenced against the Illumina-curated version of Greengenes Consortium Database. The classification step is based on ClassifyReads, a high-performance implementation of the Ribosomal Database Project (RDP) Classifier [112].

Statistical analyses

Statistical analyses applied to explore the structure of bacterial communities with respect to their diversity, composition and bacterial association patterns across sampling methods and differential taxa abundance were performed in R [92]. Graphical data analysis was performed via ggplot2 [113] and fantaxtic R packages for data visualization [110].

Alpha diversity

In order to summarize the structure of the observed bacterial communities, a set of common alpha diversity metrics in mNGS data were computed for each sample at both the species and the genus level: observed richness, Chao index [20, 21], Abundance Coverage Estimator (ACE) [25, 77], Shannon index [101], Simpson indices [54] and Fisher’s alpha index [39] using vegan [78], phyloseq [73] andPMCMRplus [88] R packages. Generalized linear mixed-effects models were constructed and ANOVA tests were computed using lme4 to analyze the effect of sampling method on alpha diversity indices [7]. Assumptions underlying parametric in parametric statistics were checked in model residuals through visual inspection (QQ Plots and density distributions) and also by significance tests (Shapiro-Wilk and Levene’s test for assessing normality and homoscedasticity). Multiple comparisons were performed via Bonferroni corrected post hoc tests in the multcomp package [52].

Data normalization

Library size was standardized across samples using different normalization approaches available in metagenomeSeq [81] and NetCoMi [87] R packages: Total Sum Scaling – TSS [18], Cummulative Sum Scaling – CSS [82] and Centered Log-Ratio transformation – CLR [2].

Differential taxa abundance

Differential abundance testing was performed at different taxonomic ranks via differential expression analyses based on multivariate differential association computed by MaAsLin2 R package [71].

Beta diversity

The effect of sampling method on beta diversity was assessed via Permutational Analyses of Variance (PERMANOVA) tests (1000 permutations) based on Aitchison distance for the community composition at both the OTU and the genus level [1]. For visual inspection, dissimilarity networks (Aitchison distance) were also constructed using NetCoMi [87] R package.

Association networks

After constructing microbial association networks based on SparCC (Sparse Correlations for Compositional data) correlation measure [40], differential network analyses were conducted using the discordant method [104]. At both the OTU and the genus levels, differential plot networks were constructed and compared using the NetCoMi R package. A sparsification threshold of 0.5 was used for comparing global network properties, centrality measures and hub taxa in the NetCoMi package via permutation tests using 1000 permutations. Adjusted Rand Index (ARI) and Graphlet Correlation Distance (GCD) measures were calculated to assess whether the clustering solutions are more or less similar than expected at random or distance similarities [91, 116]. As a measure of conditional dependence, SPIEC-EASI (Sparse InversE Covariance estimation for Ecological Association and Statistical Inference) pipeline was also applied to non-transformed data. Association network properties were computed, and keystone taxa were selected from node degree (number of connections) and betweenness (node centrality) measures using SpiecEasi R package [65].

Results

The mean age of study participants was 56 years, ranging from 38 to 71 years, whereas the sex composition was 7 females (58%) and 5 males (42%). A total of 15 implants were placed in the following positions: 14 (1), 25 (1), 26 (1), 27 (1), 34 (1), 36 (2), 37 (2), 45 (1), 46 (2) and 47 (3). The number of dental implants in which transepithelial abutments with machined surface were placed ranged from one to two (only one transepithelial abutment with machined surface was placed in a total of nine patients, while a total of two transepithelial abutments with machined surface were placed in the remaining three patients).

As shown in Fig. 1, a significant effect of sampling method on most bacterial alpha diversity indices was detected, at both the OTU level (observed [p = 0.00090], Chao index [p = 0.0026] and Fisher’s alpha index [p = 2.4E−05]) and at the genus level (observed [p = 0.019], Chao index [p = 0.047], Fisher’s alpha index [p = 0.00038] and Simpson’s indices [p = 0.014]). After performing pairwise comparisons, significant differences in observed richness, Chao and Fisher’s alpha indices were identified between ToCF/TACF and TA at both the OTU and the genus level, and in Simpson’s indices at the genus level. No significant differences were detected in Shannon or Simpson’s diversity indices at the OTU (p = 0.81 and p = 0.20, in each case) or in the Shannon index at the genus level (p = 0.30).

Fig. 1
figure 1

Boxplot of alpha diversity indices reflecting taxa abundance and consistency in samples grouped by sampling method at: A) the OTU level, and B) the genus level

Full size image

According to Fig. 2, Firmicutes (39–53%), Bacteroidota (9.4–21%), Actinobacteria (6.7–8.9%), Fusobacteria (5.9–9%) and Proteobacteria (5.5–9.5%) are the most abundant phyla in the studied samples, with a higher proportion of Firmicutes in ToCF (53%) and TACF (52%), and more Bacteroidota (21%), Fusobacteria (9.0%) and Synergistetes (1.6%) in TA samples. Considering the class level, Bacteroidia (7.4–19%) and Negativicutes (7.3-13%) were two main taxa in all cases. Nevertheless, while the proportion Actinomycetia (8.3%) was higher in ToCF and that of Bacilli was higher in both ToCF and TACF (37%), the abundance of Clostridia was higher in TACF (7.6%) and TA samples (10%). The orders Lactobacillales (12-33%), Bacteroidales (7.5-19%), Eubacteriales (7.5-10%) and Veillonellales (5.4-11%) reached the highest abundances in all sample types, together with Pasteurellales (3.5%) in TA and Actinomycetales in ToCF (5.4%) and Bacillales in TACF (4.1%) and ToCF (5.3%). On the other hand, Streptococcaceae (8.4-22%), Prevotellaceae (4.5-12%), Fusobacteriaceae (3.9-8.0%) and Veillonellaceae (5.4-11%) were the most represented families in all sample types, along with Selenomonadaceae (2.5%) in ToCF, Bifidobacteriaceae (2.3%) and Clostridiaceae (3.0%) in TACF, Enterococcaceae and Bacillaceae in TACF (9.4% and 2.5% in each case) and ToCF (11% and 2.6% in each case), and Peptostreptococcaceae (6.9%), Porphyromonadaceae (6.1%), Pasteurellaceae (3.5%) and Treponemaceae (1.2%) in TA samples. At the genus level, Streptococcus (8.3-11%) and Veillonella (4.5-6.9%) were abundant genera in all sample types. Nevertheless, ToCF and TACF samples were richer in Lactococcus (9.0% and 14% in each case) and Enterococcus (11%), whereas Prevotella (11%), Fusobacterium (8.1%), Porphyromonas (5.8%) and Peptostreptococcus (4.9%) reached higher frequencies in TA samples.

Fig. 2
figure 2

Stacked bar graphs representing cumulative abundances of taxa that represent up to 10 most abundant taxa nested by Phylum (only top 10 phyla were plotted) and grouped by sampling method (TSS normalized data). Taxonomic ranks nested by Phylum were: A) the Class level, B) the Order level, C) the Family level, and D) the Genus level

Full size image

As shown in Fig. 3, significant differences in taxa abundances between sampling methods were identified in 0.58% (10/1732) of the OTUs shared by ToCF and TA, and in 0.40% of the OTUs shared by TCFA and TA (7/1732). Significant changes were also detected after comparing TA and ToCF in 3.1% of the families (10/322), 4.1% of the orders (6/145), 8.7% of the classes (6/69) and 6.3% of the phyla (2/32); and TA and TACF in 0.60% (5/830) of the genera, 2.0% (6/306) of the families and in 1.4% (2/141) of the orders. Pairwise comparisons revealed significant changes in the frequency of some peri-implant or periodontal disease related taxa between different sampling methods. In TA samples, Synergistetes, Slackia, Peptostreptococcus, Atopobium, Mogibacterium, Slackia exigua or Peptostreptococcus stomatis showed increased abundance when compared to ToCF, whereas those of Lactococcus, Enterococcus, Cellulomonas, Corynebacterium, Exiguobacterium, Bacillus, Microbacterium and Actinomyces naturae were reduced. In contrast to TACF, TA samples also showed lower abundances of several taxa, including Bacilli, Exiguobacterium, Enterococcus, Cellulomonas, Acinetobacter, Microbacterium, Paenibacillus, Lactococcus and Actinomyces naturae.

Fig. 3
figure 3

Differentially abundant taxa between pairwise sampling method (SM) comparison identified through MaAsLin2 multivariate association testing: A) Phylum level, B) Class level, C) Order level, D) Family level, E) Genus level and F) OTU level

Full size image

Figure 4 shows compositional beta diversity biplot generated through Aitchison distance matrix considering taxa with at least 0.1% of sequencing reads. PERMANOVA analyses based on Aitchison distance showed that the effect of sampling method on bacterial community composition was statistically significant at both the OTU (p = 0.0010) and the genus level (p = 0.0010). Multilevel pairwise comparisons revealed significant differences between TA and ToCF (OTU level: p = 0.0010; genus level: p = 0.0010) and between TACF and TA (OTU level: p = 0.0020; genus level: p = 0.0030), but not between TACF and ToCF (OTU level: p = 0.837; genus level: p = 0.797).

Fig. 4
figure 4

Dissimilarity network based on Aitchison distance matrix representing the beta-diversity at: A) the OTU level, and B) the genus level retaining taxa with at least 0.1% of sequencing reads (zeros were replaced via multiplicative simple replacement and k-nearest neighbor was used as sparsification method)

Full size image

Bacterial association networks of ToCF, TACF and TA communities constructed from SparCC correlations retaining taxa with at least 0.1% of sequencing reads are represented in Fig. 5, where nodes represent bacterial taxa and edges represent either co-presence (positive association) or mutual exclusion (negative association) relationships. At both the OTU and the genus levels, pairwise network comparisons revealed high clustering similarities in terms of ARI and GCD, and no significant changes in global network properties when comparing different sampling methods (Table 1). However, significant differences were detected in terms of hub taxa between ToCF and TA samples at the OTU level (Jaccard Index = 0.00, p = 0.017), with a higher influence of Streptococcus spp., Veillonella atypica and Neisseria mucosa in ToCF, and a particular importance of Porphyromonas pasteri, Centipeda periodontii, Selenomonas sputigena, Veillonella tobetsuensis or Parvimonas micra in TA samples (Table 2). Significant differences were also identified in betweenness centrality after comparing ToCF and TA sampling methods at the OTU level (Jaccard Index = 0.13, p = 0.0033). Nevertheless, no significant differences were observed after adjusting permutation p-values of the tests for differential centrality values for multiple testing. Further information on these measures is available in Supplementary Material 2.

Fig. 5
figure 5

Bacterial association networks from ToCF, TACF and TA at the OTU and genus levels constructed using SparCC correlation coefficients for compositional data retaining taxa with at least 0.1% of sequencing reads. For network sparsification, only edges corresponding to an absolute association greater or equal than 0.7 (OTU level) and 0.6 (genus level) were represented in order to improve network readability

Full size image
Table 1 Results from testing global network metrics of the networks in Fig. 5 for group differences (1000 permutations). The computed measures for ToCF, TACF and TA, the absolute difference, and the p-value adjusted for multiple testing using the adaptative Benjamini-Hochberg method [11, 66] at both the OTU and the genus level are summarized
Full size table
Table 2 Jaccard index values (j) corresponding to Fig. 5 accounting the similarity of the sets of most central nodes (centrality value above the empirical 70% quantile) and also of the sets of hub taxa between pairs of networks (j is 0 if the sets are completely different and 1 for exactly equal sets)
Full size table

After constructing SPIEC-EASI based ecological networks, the highest rated keystone OTUs attending to their node degree and betweenness were Prevotella multisaccharivorax, Streptococcus dentapri, Peptococcus niger, Porphyromonas circumdentaria, Mycoplasma salivarium and Streptococcus gallinaceus in ToCF, Streptococcus oligofermentans, Actinomyces odontolyticus, Bacteroides heparinolyticus, Acidaminococcus intestini, Moryella indoligenes and Leptotrichia wadei in TACF, and Rothia aeria, Streptococcus sobrinus, Moryella indoligenes, Actinobaculum massiliense, Prevotella paludivivens and Megasphaera sueciensis in TA bacterial communities. At the genus level, keystone genera were Butyrivibrio, Flavobacterium, Leptotrichia, Paracoccus, Pasteurella and Sphingomonas in ToCF, Haloactinobacterium, Porphyromonas, Sphingobacterium, Shuttleworthia, Moraxella and Propionibacterium in TACF, and Bacillus, Alloprevotella, Campylobacter, Pontibacillus, Eikenella and Parascardovia in TA.

Differential network analyses (Fig. 6) revealed correlation changes between certain genera following pairwise comparisons. When comparing association patterns from TACF and TA samples, the differential network analysis revealed significant correlation changes between a large number of genera, including Clostridium and Bifidobacterium, Bifidobacterium and Dialister, Fusobacterium and Haemophilus or Actinomyces odontolyticus and Veillonella atypica (inversely correlated in TACF and no correlated in TA), and Peptostreptococcus and Haemophilus, Campylobacter and Oribacterium or Bifidobacterium and Porphyromonas (no correlated in TACF and negatively correlated in TA). Pairwise comparisons between TA and ToCF sampling methods also showed strong positive correlations in ToCF samples that remained no significant in TA samples (Fusobacterium and Selenomonas, Alkaliphilus and Dysgonomonas, Dysgonomonas and Actinomyces, Gemella and Haemophilus, Rothia and Streptococcus, Rothia and Parascardovia, or Haemophilus and Lautropia) and viceversa (Megasphaera and Prevotella). Moreover, some negative correlations between taxa detected in ToCF samples were not identified in TA samples (Treponema and Rothia, and Selenomonas and Rothia) and viceversa (Lautropia mirabilis and Haemophilus parainfluenciae, Actinomyces odontolyticus and Prevotella maculosa, or Atopobium parvulum and Streptococcus cristatus). No significant differential correlations were observed between ToCF and TACF samples.

Fig. 6
figure 6

Differential association networks based on Fig. 5: A) OTU level and B) Genus level

Full size image

The effect of sampling strategy on the metagenomic profile of transepithelial abutments at different levels of analysis, including alpha diversity, beta diversity, differential taxa abundance and association network properties, was summarized in Table 3.

Table 3 Summary of the effect of sampling method on the description of the bacterial communities from transepithelial abutments
Full size table

Discussion

When considering healthy peri-implant environments, bacterial colonization of the implant surface tends to be like healthy surrounding periodontal sites, with lower diversity [49]. However, it evolves toward the establishment of organized biofilms within the next two weeks [9]. Despite the fact that salivary pellicles adsorbed to implant surfaces promote the adhesion of microbes, its molecular features (chemical composition) and the immunological microenvironment determine the microbial colonization process [33]. In the present study, certain alpha diversity metrics (Observed taxa, Chao index, Fisher’s alpha index and Simpson’s indices) have revealed significant differences between crevicular fluid samples (ToCF and/or TACF) and transepithelial abutments (TA). Differential statistical significance between diversity metrics could be explained by the varying influence of richness, evenness and dominance in each index. Considering that TA samples, as opposed to crevicular fluid samples, account not only free-living bacteria around transepithelial abutments, but also most biofilm forming surface adhered bacteria, these results suggest that tooth surface could harbor a richer microbiota than TA surface. These results – indicating a lower diversity in peri-implant microbiota—are in agreement with those of Dabdoub et al. [28] and Payne et al. [83]. Nevertheless, bacterial diversity in peri-implantitis sites is usually higher than that of healthy periodontal sites [106, 120], but lower than that of periodontitis environments  [8].

According to the consulted bibliography, differentially abundant taxa related to healthy sites include mainly gram-positive cocci and non-motile bacilli, with higher frequencies of lactic acid bacteria (Lactobacillales and Bifidobacterium) [47], Lactococcus [15], Haemophilus [97], Veillonella [97, 102, 109], Streptococcus [97, 109], Neisseria [97], Rothia [97], Prevotella [42, 109], Actinomyces [64, 109]), Leptotrichia [15, 64], Gemella spp. [102], Vibrio [42], Brevundimonas [15, 118], Pseudomonas [118], Oribacterium spp., Selenomonas spp. and Cardiobacterium spp. [28], Staphylococcus [105], Granulicatella adjacens, Veillonella dispar, Actinomyces meyeri and Streptococcus mitis [27], Propionibacterium acnes [120], Acinetobacter, Paracoccus and Moraxella [45], Dialister [61], Abiotrophia defectiva [111], Microbacterium  [108], Corynebacterium ([105]), Novosphingobium capsulatum [36], Propionibacter, Lautropia, Chitinophagaceae, Brevundimonas nasdae, Delfitia acidivorans, Rothia aeria, Anaerofilum pentosovorans, Anaerofilum agile, Pseudoramibacter alactolyticus and Porphyromonas HMT-277/278 [15]. Moreover, Kroeger et al. [63] found higher abundances of Lautropia mirabilis, Rhodobacteriaceae or Bergeyella in shallow peri-implant pockets.

On the other hand, most published research findings confirm that peri-implantitis associated bacteria consist mainly of gram-negative motile anaerobic periopathogens and opportunistic bacteria, including higher abundancies of Porphyromonas gingivalis ([3, 10, 55, 68, 86, 97, 103, 107, 109, 117]), Aggregatibacter actinomycetemcomitans ([53, 68, 117]), Prevotella intermedia ([55, 68, 74, 109]), Prevotella nigrescens [68, 74], Treponema denticola ([17, 85, 103]), Tannerella forsythia ([3, 10, 74, 86, 103, 117]), Fusobacterium nucleatum ([3, 10, 55, 109]), Parvimonas micra [17, 62, 109], Eubacterium [64, 109], Butyrivibrio [28, 64], Filifactor alocis [97, 109], Pseudoramibacter [60], Desulfobulbus [16, 34], Streptococcus ([27, 55, 109]), Exiguobacterium [24], Streptococcus mutans and Peptococcus [28, 64]. Koyanagi et al. [62] reported that Chloroflexi, Tenericutes, Synergistetes, Peptostreptococcus stomatis and Solobacterium moorei were detected only in peri-implantitis sites.

Peri-implantitis has been also associated with higher relative frequencies of Eikenella corrodens [17], Streptococcus intermedius, Streptococcus mitis, Haemophilus influenzae and Treponema socranskii [86], Campyblobacter gracilis, Dialister invisus, Eubacterium infirmum and Mitsuokella (da Silva et al. 2013), Campylobacter rectus [17, 26], Fusobacterium [64, 74], Slackia exigua, Parascardovia denticolens and Centipeda periodontii [109], Mycoplasma and Treponema [64], Streptococcus non-mutans [28], Neisseria, Kingella, Enterococcus, Fretibacterium and Bacillus [55], Propionibacterium, Paludibacter, Staphylococcus, Filifactor and Mogibacterium [105], Treponema maltophilum [97], Olsenella and Sphingomonas [76], Veillonella [30, 64], Treponema [63, 117], Prevotella tannerae [35], Actinomyces [28], Fretibacterium fastidiosum [97], Pseudoramibacter alactolyticus [28, 62], Campylobacter [64] and Staphylococcus aureus [86], Stenotrophomonas, Leuconostoc, Faecalibacterium prausnitzii, Haemophilus parainfluenzae, Prevotella copri, Bacteroides vulgatus and Bacteroides stercoris [80]. Furthermore, the following taxa are also strongly associated with subgingival plaque of periodontitis: Alloprevotella, Phocaeicola, Johnsonella and Mycoplasma [24].

Oral micro-habitats, including teeth, transepithelial abutments and crevicular fluid, provide unique biological niches that harbors specific bacterial communities. Compositional variation in oral microbiota related to sampling origin has been highlighted by multiple studies, including crevicular fluid around dental prostheses that were fabricated by various biomaterials and fabrication techniques [13, 48], supra and subgingival biofilm [103, 117] and different peri-implantitis lesions [63, 89]. Moreover, dental intervention-related perturbations lead to significant environmental changes at the microscale level, including surface topology, chemistry and immunological response, resulting in dysbiosis and structural disruption of the oral microbiota. Compositional variation was higher between bacterial communities from transepithelial abutments (TA) and periodontal crevicular fluid (ToCF). In this sense, a significant reduction in certain peri-implant or periodontal health related taxa was observed in TA when compared to ToCF (Bacilli, Actinomycetia and Lactococcus), while several of those related with oral dysbiosis showed increased relative frequencies (Synergistetes, Peptostreptococcus, Slackia, Atopobium and Mogibacterium).

Considering higher taxonomic ranks, higher abundances of certain taxa after implant insertion (Bacteroidota, Synergistetes and Coriobacteriaceae) and lower frequencies of Actinomycetia, Bacillaceae, Cellulomonadaceae and Enterococcaceae were identified in TA samples. In this context, Heyman et al. [51] reported analogous population trends for Bacteroidota and Coriobacteriaceae after implant placement using a murine model. On the other hand, certain peri-implant and periodontal disease related taxa, including Mogibacterium, Slackia, Peptostreptococcus, Atopobium, Slackia exigua and Peptostreptococcus stomatis showed increased abundances in transepithelial abutments. Similar results were obtained after comparing crevicular fluid from transepithelial abutments (TACF) with transepithelial abutments (TA), with lower frequencies of several opportunistic commensal enterococci, Cellulomonas, Lactococcus and Actinomyces naturae in TA samples. When compared with crevicular fluid samples extracted transepithelial abutments, TA were also depleted in other taxa, including Bacilli, Paenibacillus, Exiguobacterium, Microbacterium and Corynebacterium. These results do not agree with those of Schaumann et al. [100], who concluded that peri-implant and periodontal tissues share a similar biofilm composition at the genus level.

Dental intervention-related perturbations lead to significant environmental changes at the microscale level, including surface topology and chemistry, resulting in dysbiosis and structural disruption of the oral microbiota. Abundance-based associations also revealed differential correlation across sampling methods, particularly when TA was compared to ToCF. More specifically, differential associations affecting mainly peri-implant health and peri-implantitis related taxa (Bifidobacterium, Haemophilus, Campylobacter, Peptostreptococcus, Porphyromonas, Treponema, Selenomonas, Fusobacterium, Rothia and Prevotella) were detected after performing pairwise comparisons. Moreover, significant variations regarding hub OTUs were identified between ToCF and TA. These findings suggest that both peri-implant and crevicular fluid bacterial communities are characterized by a particular distinctive associational footprint as opposed to that of the adjacent teeth or transepithelial abutments and also that environmental gradients and complex ecological interactions contribute to variations in ecosystem structure and function.

According to the results found in the present study, the abundances of most peri-implantitis or periodontal disease related taxa were higher in TA samples, whereas those of peri-implant health associated bacteria were higher in ToCF or TACF. Thus, current data suggest that implant micro-habitats are usually characterized by a dysbisotic shift, enriched in Synergistetes, Bacteroidota, Atopobium, Slackia, Peptostreptococcus, Mogibacterium and Slackia exigua, and depleted in Bacillus, Paenibacillus, Enterococcus, Lactococcus, Microbacterium, Corynebacterium and Actinomyces naturae. Differential abundance patterns between crevicular fluid from transepithelial abutments and transepithelial abutments, not detected when comparing crevicular fluid samples from different origins, reflect the particularities of specific oral niches [69]. Furthermore, sampling method was a significant source of bacterial community dissimilarity. In this sense, significant differences were identified at the community level between TA and ToCF and also between TACF and TA, but not between ToCF and TACF. With respect to network properties, sampling procedures did not significantly differ in terms of global network properties or centrality measures. However, variations in hub OTUs were identified between ToCF and TA. By contrast, TACF and TA shared the most relevant hub taxa at both the OTU and the genus level.

As stated by Berry and Widder [12], hub taxa are not necessarily keystones in the microbial community, but ecologically relevant hub taxa are likely to be. Thus, different keystone taxa were also identified in bacterial communities from different sampling origins. Considering node degree and betweenness, Streptococcus oligofermentans, Actinomyces odontolyticus, Bacteroides heparinolyticus, Acidaminococcus intestini, Moryella indoligenes, Leptotrichia wadei, Haloactinobacterium, Porphyromonas, Sphingobacterium, Shuttleworthia, Moraxella and Propionibacterium have a pivotal role in TACF samples, whereas Rothia aeria, Streptococcus sobrinus, Moryella indoligenes, Actinobaculum massiliense, Prevotella paludivivens, Megasphaera sueciensis, Bacillus, Alloprevotella, Campylobacter, Pontibacillus, Eikenella and Parascardovia showed a higher influence in TA samples. On the other hand, Prevotella multisaccharivorax, Streptococcus dentapri, Peptococcus niger, Porphyromonas circumdentaria, Mycoplasma salivarium, Streptococcus gallinaceus, Butyrivibrio, Flavobacterium, Leptotrichia, Paracoccus, Pasteurella and Sphingomonas were identified as drivers of microbiome structure and functioning in ToCF samples. With the exception of Flavobacterium (ToCF), Sphingobacterium (TACF), Rothia aeria (TA), Moraxella (TACF), Sphingomonas (ToCF) and Rothia mucilaginosa (TA), most keystone taxa were obligate (eg. Bacteroides heparinolyticus, Moryella indoligenes, Acidaminococcus intestine, Peptococcus niger, Megasphaera sueciensis, Shuttleworthia, Campylobacter, Butyrivibrio, Actinomyces, Propionibacterium) or facultative anaerobes (eg. Streptococcus spp., Parascardovia, Alloprevotella, Bacillus, Paracoccus, Leptotrichia, Eikenella, Porphyromonas, Pasteurella, Mycoplasma salivarium or Actinobaculum massiliense).

Conclusions

According to these findings, it should be highlighted that bacterial community assessment via mNGS could be considered a promising strategy for peri-implant and periodontal health surveillance and early peri-implant disease diagnosis. Despite the fact that, regardless of sampling approach, association network properties and centrality measures from transepithelial bacterial communities were similar, significant differences were detected in terms of hub taxa, alpha and beta diversity and individual taxa abundance, including some peri-implant and periodontal health related taxa. In most aspects, crevicular fluid samples (ToCF and TACF) are not representative of bacterial communities developed on transepithelial surfaces. As a result, it can be concluded that the choice of sampling strategy can deeply affect the results of oral microbiota profiling, with a particular emphasis on peri-implant health related bacterial taxa.