Apical periodontitis: preliminary assessment of microbiota by 16S rRNA high throughput amplicon target sequencing

The study was planned and performed in accordance with the Declaration of Helsinki and was approved by the Ethics committee of the Dental School, University of Turin. From January 2015 to February 2016, 121 patients with apical periodontitis were referred to the Triage of the Dental School of the University of Turin. Complete medical history and panoramic radiograms (OPTs) of each patient were assessed seeking large Apical periodontits (APs) that clearly had developed in proximity to teeth with bad prognosis. In the Oral Surgery Department, 37 patients were selected by applying the following exclusion criteria: systemic or local disease or condition (hematologic diseases, uncontrolled diabetes, serious coagulopathies, history of intravenous therapy with bisphosphonates, and/or diseases of the immune system) possibly precluding oral surgical intervention; immunosuppression; HIV+, HCV+, HBV+, TBC+, corticosteroid treatment, pregnancy, radiotherapy to the head or neck region within 12 months before surgery. A further restriction was achieved excluding teeth a) periodontally compromised, b) with endodontic communication to the oral cavity and c) vertical tooth fractures. Finally, after giving their informed consent, 10 patients underwent surgery. By careful manipulation, the periapical lesions were harvested sterilely and fixed in 4% formalin, while the hopeless teeth were extracted. Based on histology and a dimensional cut-off (lesion greater than 1 cm in diameter), among the apical periodontal lesions analyzed, 5 Radicular Cysts (RCs) and 5 Periapical Granulomas (PGs) were retrieved, dependently on the clear presence or absence of cavity lining epithelium, respectively.

The histological specimens retrieved were fixed in 4% formalin for 24 h and subsequently embedded with paraffin wax and cut into 3 μm thick sections, using a motorized microtome. Polylysine coated slides were used to enhance the adhesion of the tissue section during staining procedures. The histological structure of the lesions was assessed by traditional hematoxylin and eosin staining for optical microscopy.

Two slices of formalin-fixed, paraffin-embedded (FFPE) tissue samples (about 10 mg of tissue) were used for total genomic DNA extraction. Samples were pre-treated at 55 °C for a minimum of 1 h with the dissolving buffer and Proteinase K (20 mg/ml) according to the manufacturer’s instructions (BiOstic® FFPE Tissue DNA Isolation Kit Mobio, Carlsbad, CA, USA). DNA was used to study the microbial diversity by pyrosequencing of the amplified V1–V3 region of the 16S rRNA gene, recurring to the primers Gray28F (5’-TTTGATCNTGGCTCAG) and Gray519r (5’-GTNTTACNGCGGCKGCTG) that amplify a fragment of 520 bp, following PCR conditions previously reported [17]. PCR products were purified twice with Agencourt AMPure purification kit (Beckman Coulter, Milan, Italy), and quantified using the PlateReader AF2200 (Eppendorf, Hamburg, Germany) with PicoGreen assay and an equimolar pool was obtained prior to further processing. Due to poor DNA quality, PG_8 sample was excluded. The amplicon pool was used for pyrosequencing on a GS Junior platform (454 Life Sciences, Roche, Monza, Italy) according to the manufacturer’s instructions by using Titanium chemistry.

QIIME 1.9.0 software was employed to analyze 16S rRNA data [18]: OTUs (operational taxonomic units) were picked at 99% of similarity by means of UCLUST clustering methods [19]. Representative sequences from each cluster were used to assign taxonomy through matching against the Greengenes 16S rRNA gene database version 2013 by the RDP classifier. R environment (www.​r-project.​org) was adopted to elaborate statistics as well as plotting. To calculate the microbiota alpha diversity the authors chose the “diversity” function of the vegan package of R. Weighted UniFrac distance matrices as obtained through QIIME were imported in R to generate PCoA (Principal Coordinates Analysis) plots. Weighted UniFrac distance matrices were also used to perform ADONIS and ANOSIM statistical tests owing to compare_categories.py script of QIIME. OTU tables, which were filtered at 0.2% abundance in at least two samples, were used to compare each OTU based on the passed sample groupings (PGs and RCs) through the group_significance.py script of QIIME. OTUs co-occurrence co-exclusion was carried out by the psych package of R (www.​r-project.​org) and it was further visualized through the corrplot package of R [20]. In order to predict the inferred metagenome, the authors used PICRUSt [21] so as to predict abundances of gene families based on 16S rRNA sequences data [22]. Briefly, the pick OTUs step was re-performed at 97% similarity against the Greengenes database and the resulting KEGG orthologs table was then collapsed at level 3 of the KEGG annotations in order to display the inferred metabolic pathways. The resulting table was imported in the GAGE Bioconductor package [23] to identify biological pathways overrepresented or underrepresented between PGs and RCs samples. To characterize the accuracy of PICRUSt, the Nearest Sequenced Taxon Indexes (NSTI) were calculated [21]. All the sequencing data were deposited at the Sequence Read Archive of the National Center for Biotechnology Information (accession numberSRP096711).

A total of 163.832 raw reads were obtained after the sequencing. 65.070 reads passed the filters applied through QIIME, with an average value of 7.230 reads/sample and a sequence length of 486 bp. The rarefaction analysis and the estimated sample coverage (Table 2) indicated that there was a satisfactory coverage for all the samples (ESC > 96%). The richness of the samples varied from a minimum of 22 to a maximum of 202 OTUs. (Table 2) Alpha-diversity indices (Table 2) showed no difference on the level of complexity (P > 0.05) of RCs samples compared to PGs.

In Fig. 1, the box plot shows the OTUs with a relative abundance of 0.2% in at least two samples (Fig. 1). The data showed a varied microbiota composition characterized by the presence of 40% of facultative microbes , 37% of anaerobes and 22% of aerobe microbes (Table 3). In details, the samples were characterized by the predominance of Lactococcus lactis (55% of the relative abundance), Propionibacterium acnes (18%), Corynebacterium matruchotii (5.5%), Staphylococcus warneri (5%), Gemellales (2%), Actinomyces johnsonii (2%), and Lactobacillus zeae (2.5%). Through principal coordinate analysis (PCoA) with a weighted UniFrac distance matrix, it was possible to show that samples from RCs grouped together and that they were well separated from PGs on the basis of their microbiota (Fig. 2); ADONIS and ANOSIM statistical tests confirmed this difference (P < 0.001). The differential abundance analysis showed a higher abundance (Bonferroni corrected P value of < 0.001) of several minor OTUs in RCs compared to PGs samples. In particular, it was possible to observe the most abundant presence of several facultative anaerobes or anaerobe OTUs such P. acnes, Gemellales, Capnocytophaga ochracea, Paracoccus, Fusobacterium nucleatum, Prevotella intermedia and Rothia dentocariosa. However L. lactis was found significantly more abundant (P < 0.001) in PGs samples than in RCs samples.

The OTU co-occurrence was investigated by considering the species-level taxonomic assignment and significant correlations at false-discovery rate [FDR], < 0.05. (Fig. 3). P. intermedia showed the highest number of positive correlations including those with Streptococcus mitis and F. nucleatum and a co-exclusion with Sphyngomonas sp. Moreover the most significant OTUs in cyst samples such as Gemellales, C. ochracea, showed the highest number of positive correlation. The core OTUs L. lactis co-exclude the presence of Acinetobacter lwoffii, while P. acnes co-exclude the presence of S. mitis. (Fig. 3).

Regarding the predicted metagenomes, the weighted nearest-sequenced-taxon index (NSTI) for the samples, expressed as the mean SD, was 0.042 ± 0.004. Thus, a NSTI score of 0.042 indicates a satisfactory accuracy for all of the samples (96%). The pathway enrichment analysis (performed by GAGE) of the predicted metagenomes showed an enrichment of metabolic pathways such as Biosynthesis of amino acids (ko01230), Pyruvate metabolism (ko00620), Propanoate metabolism (ko00640) in PGs samples compared to RCs samples (data not shown). In contrast, from RCs samples only pathways involved in cellular processes, biosynthesis of secondary metabolites, and genes involved in Lipopolysaccharide biosynthesis (ko00540) were found.