Metabolic phenotyping of saliva to identify possible biomarkers of periodontitis using proton nuclear magnetic resonance

Seonghye Kim1|,Hyun-Joo Kim,|,Yuri Song4,5,|,Hyun Ah. Lee4,5,| Suhkmann Kim1|,Jin Chung4,5

Abstract

Aim: The aim of this study was to propose biomarker candidates for periodontitis via untargeted metabolomics analysis.
Materials and methods: Metabolic profiling was performed using saliva samples from 92 healthy controls (H) and 129 periodontitis patients (P) in the discovery cohort using proton nuclear magnetic resonance spectroscopy. Random forest was applied to identify metabolites that significantly differentiated the control group from the periodontitis group. Candidate metabolites were then validated in an independent validation cohort.
Results: In the discovery set, the metabolic profiles of the P group were clearly separated from those of the H group. A total of 31 metabolites were identified in saliva, and 7 metabolites were selected as candidate biomarkers. These metabolites were further confirmed in the validation set. Ethanol, taurine, isovalerate, butyrate, and glucose were finally confirmed as biomarkers. Furthermore, the biomarker panel showed more than 0.9 of the area under curve value in both discovery and validation sets, indicating that panels were more effective than individual metabolites for diagnosing periodontitis.
Conclusions: We identified five metabolite biomarkers that discriminated patients with periodontitis from healthy controls in two independent cohorts. These biomarkers have the potential for periodontal screening, detection of periodontitis, and monitoring of the outcome of periodontal therapy.

K E Y W O R D S
diagnosis, metabolomics, nuclear magnetic resonance spectroscopy, periodontitis, saliva

1,|,INTRODUCTION

Periodontitis is a dysbiotic, bacterium-driven inflammatory disease that affects periodontal tissues. Its severity ranges from mild, which is confined to the gingiva, to severe, which causes extensive loss of bone and tooth (Hajishengallis, 2015). Until now, the most widely used method for the diagnosis of periodontitis is the measurement of the clinical attachment level (CAL) and periodontal probing depth (PD) using a periodontal probe and checking for alveolar bone loss on a radiograph (Tonetti et al., 2018). However, this traditional diagnostic procedure has several limitations: it requires highly trained experts and assistants to measure and record the related clinical parameters; it is difficult to obtain highly accurate agreement between examiners; the probing may cause discomfort to patients; and it is both time consuming and costly (Miller et al., 2010). Moreover, diagnosis based on clinical findings and radiographs can be confirmed only after periodontitis has progressed to some extent, and because of this, only the results of past disease progression can be confirmed but not the current disease activity. Therefore, in recent years, diagnosis of periodontitis through the discovery of biomarkers at the molecular level has come into the spotlight. Saliva, which can be easily obtained, is actively used because it can reflect the pathological molecular change of periodontitis immediately at the patient level (Kc et al., 2020). Salivary metabolomics targets small molecules derived from the dysbiotic microbiome and host tissue destruction. The metabolic profile of saliva can reflect the real-time molecular phenotype of oral health; therefore, metabolomics is a rapidly evolving technique to discover biomarkers for the diagnosis and prognosis for oral diseases (Buzalaf et al., 2020). Nuclear magnetic resonance (NMR) is a robust and quantitative analysis technique that has advantages such as non-invasive metabolite detection and high reproducibility (Takeda et al., 2009). Several studies of NMR-based metabolomics have provided promising preliminary results related to the discriminatory metabolites found in periodontitis (Aimetti et al., 2011; Rzeznik et al., 2017; Romano et al., 2018), but these still need to be verified and confirmed via further research using larger sample sizes.

Clinical Relevance

Scientific rationale for study: In the development of periodontitis, the interaction between periodontal pathogens and the host affects their respective metabolic activities, leading to changes in the content of salivary metabolites.
Principal findings: 1H NMR-based metabolomic analysis identified and validated salivary biomarkers for periodontitis with large cohorts. The combined use of multiple statistical methods provided more comprehensive information for biomarker selection. Ethanol, taurine, isovalerate, butyrate, and glucose were discovered as salivary biomarkers with good capability to distinguish periodontitis.
Practical implications: Detecting selected salivary metabolite biomarkers could be developed to a simple clinical test for the detection, diagnosis, or monitoring of periodontitis.
This study was designed to differentiate between the biochemical signatures of healthy patients (H) and those with periodontitis (P) via untargeted metabolomic analysis using 1H NMR spectroscopy and thereby determine and evaluate potential biomarkers for periodontitis. Furthermore, the identified discriminant metabolites of the results were validated to confirm their potential for predicting periodontitis.

2,|,MATERIALS AND METHODS

2.1,|,Study population and assessment of periodontitis

We designed a two-stage cohort, with one cohort for biomarker discovery and an independent cohort for validation. Subjects were recruited at the Department of Periodontics, Pusan National University Dental Hospital (Yangsan, Korea).
For clinical examination, medical and dental histories were obtained. We excluded individuals using the following exclusion criteria: (1) those who had systemic diseases that might affect periodontal status, and those who were pregnant or breastfeeding; (2) those who had taken systemic antibiotics, anti-inflammatory drugs, or oral antiseptic agents within 6 months, or had received periodontal therapy (scaling and root planing) in the last 3 months; (3) those who had an acute infection (e.g., herpetic gingivostomatitis) or chronic mucosal lesions (e.g., pemphigus, pemphigoid) of the oral cavity; (4) those who had less than 20 teeth; (5) those who were smokers; and (6) those who did not agree to participate in the research and sign an informed consent.
Periodontal diagnosis was performed according to the guidelines presented by the 2017 World Workshop on the Classification of Periodontal and Peri-implant Diseases and Conditions (Caton et al., 2018). Subjects without detectable inter-dental clinical attachment loss were diagnosed as free of periodontitis and placed in the healthy group. The discovery cohort included subjects with moderate (stage II) or severe (stage III) periodontitis. Staging relied on the severity, which was classified according to the following criteria: (1) stage II, (a) interdental clinical attachment loss: 3–4 mm; (b) radiographic bone loss extending between 15% and 33% of the root length; (c) maximum PD ≤ 5 mm; and (d) no tooth loss due to periodontitis; and (2) stage III, (a) interdental clinical attachment loss ≥5 mm; (b) radiographic bone loss extending to the mid-third of the root and beyond; (c) maximum PD ≥ 6 mm; and (d) tooth loss due to periodontitis of ≤4 teeth. Patients with grade C, which indicates a rapid rate of progression or early onset disease, were excluded from this study. Periodontal status was evaluated by one periodontal specialist who recorded the following parameters: plaque index (PI), PD, CAL, and bleeding on probing (six sites per tooth).

2.2,|,Saliva sampling

Saliva samples were taken after informing the patient on the sampling protocol and before further periodontal clinical examinations were performed. All subjects were asked to refrain from consuming food or drinks, brushing, or using mouth wash for at least 1 h before sampling, and were scheduled for saliva sampling between 9:00 AM and 11:00 AM. Stimulated whole saliva was collected using a hygienic collection system (Salivette; Sarstedt, Nümbrecht, Germany). A plain cotton roll was placed in the mouth of the subject for approximately 1 min to stimulate salivation. We did not use paraffin, which can affect salivary metabolites (Figueira et al., 2017). The cotton roll, following the absorption of a sufficient amount of saliva, was placed into a Salivette tube and immediately transported to the laboratory. The cotton roll was centrifuged at 1000g for 20 min at 4C, and the supernatant was removed and stored at 80C.

2.3,|,Sample preparation and NMR measurement

For NMR analysis, the saliva samples were thawed and centrifuged at 10,000 rpm for 1 min. The supernatant (450 μl) was added to 50 μl of phosphate buffer (pH 7.4) in deuterated water (D2O, 99.9% D) containing 20 mM 3-(trimethylsilyl) propionic-2,2,3,3-d4 acid sodium salt (TSP-d4). Each sample was transferred into a 5-mm NMR tube. All H-NMR spectra were measured using a 600-MHz NMR spectrometer (Agilent, Santa Clara, CA) operating at 25C. The spectral acquisition was based on a Carr–Purcell–Meiboom–Gill pulse sequence for the suppression of water and macromolecular peaks. Spectra were measured using the following parameters: a 9.8 μs 90 pulse, a relaxation delay of 3 s, an acquisition time of 3 s, and a total acquisition time of 13 min and 9 s. A total of 128 scans were acquired at a spectral width of 24,038.5 Hz. The acquired spectra were manually phased and baselinecorrected using the VnmrJ 4.2 software (Aligent Technologies).

2.4,|,Metabolite identification and statistical analysis

Chenomx NMR suite 8.4 (Chenomx Inc., Edmonton, Canada) was used for the qualitative and quantitative analyses of the metabolites. For pre-processing the NMR spectra, each spectrum was binned from 0.5 to 7.5 ppm with a binning size of 0.003 after excluding the water region (4.25–5.5 ppm), and the total area normalization was performed. All binning data were imported into the SIMCA-P+ 12.0 software (Umetrics, Umeå, Sweden), and principal component analysis (PCA) and supervised orthogonal partial least squares discriminant analysis (OPLS-DA) were performed after scaling the unit variance differences. The models were evaluated using two parameters: R2 (the goodness of fit) and Q2 (the predictive ability). In OPLS-DA, metabolites with variable importance for projection (VIP) scores larger than 2 were considered to significantly contribute to sample separation.
The quantified metabolites were analysed using the t-test adjusted to a false discovery rate (FDR) based on the Benjamini– Hochberg and one-way analysis of variance with the post hoc test (Tukey’s HSD). The significance threshold was set as 0.05. Random forest (RF) was applied to identify candidate biomarkers using MetaboAnalyst 5.0 (Xia Lab, McGill University, Montreal, Canada). For analysis, 1000 trees were used for building the classifier, and the outof-bag (OOB) error was obtained to estimate of the classification error. Variables were ranked by their contributions to classification accuracy (mean decrease accuracy [MDA]). For the identified biomarkers, multiple linear regression, with stepwise selection of the independent variables, was performed to assess the association between clinical parameters and metabolite biomarkers using SPSS 25 (SPSS, Inc., Chicago, IL). To evaluate the diagnostic performance of selected metabolites, receiver operating characteristics (ROCs) were obtained using MetaboAnalyst 5.0.

3,|,RESULTS

3.1,|,Characterization of the subjects into the discovery and validation cohorts

A total of 271 subjects were recruited for this study. In the discovery phase, the analysis was performed on 92 healthy controls and 129 patients with periodontitis. Patients with periodontitis were divided into two subgroups: 67 patients with stage II periodontitis and 62 with stage III. For validation, saliva samples from 17 controls and 33 periodontitis patients were analysed. The demographic and clinical characteristics of the subjects are presented in Table 1. We matched the groups for gender (p > .05). However, the mean age in the H group was significantly less than in the P groups, whereas they were similar for stage II and stage III of periodontitis. The clinical parameters (PD, CAL, GI, and PI) of the P group were significantly higher than those of the H group in both cohorts.

3.2,|,Salivary metabolic profiling

A representative 1H-NMR spectrum of saliva with assigned metabolites is shown in Figure 1. A total of 31 metabolites were identified and quantified. All metabolites were verified as reported in the literature (Aimetti et al., 2011; Dame et al., 2015) and the Human Metabolome Data Base. Additionally, single spectral areas and overlapping spectral areas were confirmed by 2D correlation spectroscopy NMR spectra. The quantified concentrations of metabolites in the H and P groups are shown in Table S1.
There were significant differences in age between healthy controls and patients with periodontitis (p < .001). To assess the effect of age on the metabolic profiles in saliva, the H and P groups, including both cohorts, were divided into subgroups by age with a step of 10 years. We confirmed that there was no discrimination in the results of multivariate analysis (Q2 < 0, Figure S1) and no significant differences in metabolite concentrations among age-subgroups (p > .05, Table S2). This indicates that age differences did not affect the metabolite profile of saliva in this study.

3.3,|,Discovery of candidate biomarkers in periodontitis

Utilizing the NMR spectrum from the discovery set, PCA was performed to get an overview of the salivary metabolic pattern. In the PCA score plots, samples of the P group were scattered and separated from those of the H group (R2X = 0.697 and Q2 = 0.351; Figure 2a). OPLS-DA was processed for discrimination between the groups. The OPLS-DA score plot showed a clear separation (R2Y = 0.852 and Q2= 0.721; Figure 2b). In the VIP plot of OPLS-DA, the most important variables that contributed to the separation were identified as taurine (3.329 and 3.42 ppm) and glucose (3.261 and 3.408 ppm) (Figure 2c). OPLS-DA was also performed between the stage II and stage III groups; however, the results were not significant (Q2 < 0, Figure S2).
The 19 metabolites showed significant changes (FDR-adjusted p < .05). However, no significant difference was observed between the stage II and stage III groups.
For the biomarker selection, RF analysis was conducted with the identified metabolites in the discovery set. RF showed a significant classification between the H and P groups with an OOB error of 0.0814, and the top-ranked eight metabolites with an MDA > 0.01 were selected as candidate biomarkers (Figure 3a). ROC curve analysis was performed on these metabolites, imposing an area under curve (AUC) cut-off value of 0.7 (Mandrekar, 2010). The AUC values of seven metabolites, except propionate, showed good predictability (Table 2).

3.4,|,Validation and evaluation of biomarkers

A validation study was performed with an independent validation cohort using the same criteria. As in the discovery set, the PCA and OPLS-DA score plots showed distinct separation between groups of the validation set (Figure S3). In the results of RF for the validation data, five metabolites, namely ethanol, taurine, glucose, butyrate, and isovalerate, were confirmed as significant with an OOB error of 0.22 (Figure 3b). Two metabolites, acetone and lysine, were not significantly different in the validation set, and they were also not validated in RF results (p < .05). Therefore, five metabolites were finally selected as the biomarkers for periodontitis; the levels of these biomarkers in both the discovery and the validation set are shown in Figure 4. Furthermore, multiple regression analysis showed a relationship between each biomarker and the clinical parameters (Table S3), suggesting that these metabolites were significantly associated with one or more periodontitis indexes.
Multivariate ROC analysis was performed to evaluate the diagnostic ability of the selected biomarker panel. The results showed perfect diagnostic performance, with AUC values of 0.949 in the discovery set (87.0%, specificity 91.5%, and accuracy 89.6%; Figure 5a) and 0.906 in the validation set (sensitivity 82.4%, specificity 84.8%, and accuracy 80.4%; Figure 5b). Furthermore, the AUC value was 0.954 for all subjects from both the discovery and validation cohorts (sensitivity 86.2%, specificity 90.7%, and accuracy 89.6%; Figure 5c). These ROC models for the panel of metabolites were validated by a permutation test, which demonstrated high statistical significance (p < .002, Figure S4).

4,|,DISCUSSION

The purpose of our study was to apply high-throughput metabolomics to analyse saliva samples to identify differentiating periodontitisrelated metabolites and thereby enable biomarker discovery and validation. It has already been demonstrated that salivary metabolites are not affected by clinical features (gender or body mass index) and dietary intake (Walsh et al., 2006; Bertram et al., 2009). In addition, we confirmed that there was no marked effect of age and excluded the influence of age as a confounding factor in this study.
Interestingly, we did not find any clear a metabolic difference between stage II/stage III groups following their classification based on disease stage. Diagnosis of stages II and III of periodontitis is primarily based on clinical examination and radiographic parameters (Kornman & Papapanou, 2020), which measure only the result of periodontal tissue destruction and do not reflect current disease activity. Molecular-level analysis related to periodontitis, such as metabolites, may not follow the stage of periodontitis based on clinical parameters; rather, they may be indicators of disease activity. Therefore, the change in metabolite level due to periodontitis may act as a measure different from traditional clinical parameters, and it has been proven in previous studies that various metabolites do not necessarily show a linear relationship with the severity of periodontitis (Waddington et al., 1996; Huri et al., 2003). Therefore, identification of potential metabolite biomarkers was performed without considering disease severity according to our results.
We performed both discovery and validation for biomarkers, and RF was used for both classification and biomarker selection. This analysis is a powerful non-parametric classification method for metabolomic data processing (T. Chen et al., 2013). The candidate metabolites selected in RF were evaluated for diagnostic potential using AUC values. In our study, five metabolites (ethanol, taurine, isovalerate, butyrate, and glucose) were finally identified; these showed significant association with periodontal parameters and the greatest performance for periodontitis diagnosis in both the discovery and validation sets, either independently or in the panel.
Among the identified biomarkers, the concentrations of taurine, isovalerate, butyrate, and glucose were significantly increased in periodontitis patients in both cohorts compared with that in healthy controls. Most NMR-based metabolomic analyses have shown that butyrate and isovalerate, the major short-chain fatty acids (SCFAs), were up-regulated in patients with periodontitis (Baima et al., 2021). In particular, it was demonstrated that butyrate increased with the progress of the periodontitis stage. Many previous studies have confirmed that these metabolites are end-products of bacterial metabolism and play an important role in periodontal disorders. Bacterial metabolism in periodontal disease results in the deamination of nitrogen compounds, proteins, and amino acids and their conversion to SCFAs to produce energy (Takahashi, 2015). Also, SCFAs are associated with periodontal inflammation because they have a strong correlation with deep probing depth, loss of attachment, and bleeding on probing (Rzeznik et al., 2017). The concentrations of butyrate and isovalerate in gingival crevicular fluid are known to gradually increase with the recolonization of periodontal pathogens (Qiqiang et al., 2012). Furthermore, butyrate and isovalerate in the saliva have been found to be significantly higher at sites positive for Porphyromonas gingivalis, which we previously identified as a significant periodontitis-associated bacterial species (Na et al., 2020). Therefore, the increase in the levels of butyrate and isovalerate is an important indicator of the growth of pathogenic subgingival microorganisms and progression of periodontal tissue destruction.
The increase in bacterial metabolic and cytotoxic end-products can induce oxidative stress and host defence responses, which can enhance the growth of periodontal pathogenic bacteria such as P. gingivalis and oral Treponema (Takahashi, 2015). Taurine, whose primary role in the immune system is related to its anti-oxidant effect, is known to accumulate in inflammatory lesions (Kim & Cha, 2014). Taurine can protect tissues and maintain cellular homeostasis under conditions of inflammation and oxidative stress (Marcinkiewicz & Kontny, 2014). Recently, changes in the levels of taurine were demonstrated to constitute a protective effect mediated by the regeneration of inflammatory and anti-oxidant properties in periodontal tissues (Gawron et al., 2019). Taurine is transformed to taurine N-chloramine (TauCl), which specifically modulates the inflammatory response within the periodontal tissues. In defence reactions to bacteria, TauCl reduces the production of pro-inflammatory cytokines (IL-1β, IL-6, and IL-8) and reactive oxygen species in periodontitis (Mainnemare et al., 2004). Here, taurine was associated with the excellent diagnostic performance, with an AUC of more than 0.8 in the validation set; the taurine levels were significantly higher in periodontitis patients than in the controls. This may be the result of the anti-oxidant properties and immunomodulatory effects, serving as a host defence mechanism in periodontitis.
Moreover, P. gingivalis and many other bacteria are unable to utilize saccharides as energy sources, although amylase activity and monosaccharide levels are increased during periodontitis (Barnes et al., 2011). Salivary amylase hydrolyses starch to glucose, and amylase activity increases by binding proteins in bacteria. Thus, high levels of glucose and oligosaccharides indicate excessive amylase activity in periodontitis. Furthermore, it is interesting to note that the uptake of saccharides for energy is more active during an aggressive form of periodontitis (H. W. Chen et al., 2018). Although the other saccharides and dipeptides were not detected, the elevated glucose concentration in our study indicates that the energy demand of pathogenic bacteria was increased and a more favourable energy environment for oral bacteria was generated during periodontitis.
Conversely, the concentration of ethanol significantly decreased in periodontitis patients compared to controls, and this showed the highest diagnostic performance in both cohorts. The same trend in periodontitis has been observed previously (Rzeznik et al., 2017; Gawron et al., 2019). Ethanol is a volatile organic compound in the saliva, produced by bacterial alcohol dehydrogenase or from alcoholic beverages and can be converted to acetaldehyde, which is known to be carcinogenic. The bacterial production of acetaldehyde from ethanol in saliva is linked with poor dental status and may contribute to oral cancer (Homann et al., 2001). Since the subjects of this study showed strict limitations with regard to alcohol intake, changes in ethanol would be the result of bacterial metabolism. In addition, although acetaldehyde was not detected by NMR, the decrease in ethanol concentration was more pronounced in periodontitis patients, suggesting that the oxidation of ethanol was activated to a great extent by periodontitis.
Despite the good diagnostic performance in both the discovery and validation sets, acetone did not show a significant change in the validation set. Acetone is the byproduct of acetoacetate breakdown and is related to halitosis and poor oral hygiene. The metabolism of acetone can produce glucose via lactate, which was increased in our study (Kalapos, 2003). Also, the decrease of acetone has been confirmed in small-scale metabolomics analysis using NMR (Gawron et al., 2019). Although the results in this study were reliable and obtained using a sufficient number of samples and statistical analyses, further pre-clinical studies are required to confirm our findings. Further studies are considered essential to comprehensively reveal the association between acetone levels in saliva and periodontitis.

5,|,CONCLUSION

We performed this study to reveal the feasibility of using salivary metabolomic analysis for the detection of periodontitis. The combined multiple methods, including RF analyses, t-tests, and ROC curves, provided highly reliable and comprehensive information. We discovered and validated five metabolite biomarkers for periodontitis, namely ethanol, taurine, isovalerate, butyrate, and glucose.
Further prospective longitudinal studies are essential to comprehensively validate and confirm these biomarkers using larger cohorts. Our results demonstrate that NMR-based metabolomic studies in saliva can be applied for further clinical diagnosis of periodontitis. In conclusion, we suggest that these salivary metabolite biomarkers are valuable for periodontal screening, detection of periodontitis, and monitoring the outcome of periodontal therapy.

REFERENCES

Aimetti, M., Cacciatore, S., Graziano, A., & Tenori, L. (2011). Metabonomic analysis of saliva reveals generalized chronic periodontitis signature. Metabolomics, 8(3), 465–474. https://doi.org/10.1007/s11306-0110331-2
Baima, G., Iaderosa, G., Citterio, F., Grossi, S., Romano, F., Berta, G. N., Buduneli, N., & Aimetti, M. (2021). Salivary metabolomics for the diagnosis of periodontal diseases: A systematic review with methodological quality assessment. Metabolomics, 17, 1. https://doi.org/10.1007/ s11306-020-01754-3
Barnes, V. M., Ciancio, S. G., Shibly, O., Xu, T., Devizio, W., Trivedi, H. M., Guo, L., & Jönsson, T. J. (2011). Metabolomics reveals elevated macromolecular degradation in periodontal disease. Journal of Dental Research, 90(11), 1293–1297. https://doi.org/10.1177/ 0022034511416240
Bertram, H. C., Eggers, N., & Eller, N. (2009). Potential of human saliva for nuclear magnetic resonance-based metabolomics and for healthrelated biomarker identification. Analytical Chemistry, 81(21), 9188– 9193. https://doi.org/10.1021/ac9020598
Buzalaf, M. A. R., Ortiz, A. D., Carvalho, T. S., Fideles, S. O. M., Araujo, T. T., Moraes, S. M., Buzalaf, N. R., & Reis, F. N. (2020). Saliva as a diagnostic tool for dental caries, periodontal disease and cancer: Is there a need for more biomarkers? Expert Review of Molecular Diagnostics, 20(5), 543–555. https://doi.org/10.1080/14737159.2020. 1743686
Caton, J. G., Armitage, G., Berglundh, T., Chapple, I. L. C., Jepsen, S., Kornman, K. S., Mealey, B. L., Papapanou, P. N., Sanz, M., & Tonetti, M. S. (2018). A new classification scheme for periodontal and peri-implant diseases and conditions - Introduction and key changes from the 1999 classification. Journal of Clinical Periodontology, 45(Suppl 20), S1–s8. https://doi.org/10.1111/jcpe.12935
Chen, H. W., Zhou, W., Liao, Y., Hu, S. C., Chen, T. L., & Song, Z. C. (2018). Analysis of metabolic profiles of generalized aggressive periodontitis. Journal of Periodontal Research, 53(5), 894–901. https://doi.org/10. 1111/jre.12579
Chen, T., Cao, Y., Zhang, Y., Liu, J., Bao, Y., Wang, C., Jia, W., & Zhao, A. (2013). Random forest in clinical metabolomics for phenotypic discrimination and biomarker selection. Evidence-based Complementary and Alternative Medicine, 2013, 298183. https://doi.org/10.1155/2013/ 298183
Dame,,Z.,T.,,Aziat,,F.,,Mandal,,R.,,Krishnamurthy,,R.,,Bouatra,,S., Borzouie, S., Guo, A. C., Sajed, T., Deng, L., Lin, H., Liu, P., Dong, E., & Wishart, D. S. (2015). The human saliva metabolome. Metabolomics, 11(6), 1864–1883. https://doi.org/10.1007/s11306-015-0840-5
Figueira, J., Gouveia-Figueira, S., Ohman, C., Lif Holgerson, P., Nording, M. L., & Ohman, A. (2017). Metabolite quantification by NMR and LC-MS/MS reveals differences between unstimulated, stimulated, and pure parotid saliva. Journal of Pharmaceutical and Biomedical Analysis, 140, 295–300. https://doi.org/10.1016/j.jpba.2017. 03.037
Gawron, K., Wojtowicz, W., Łazarz-Bartyzel, K., Łamasz, A., Qasem, B., Mydel, P., Gajewska, M. C., Potempa, J., & Mlynarz, P. (2019). Metabolomic status of the oral cavity in chronic periodontitis. In Vivo, 33(4), 1165–1174. https://doi.org/10.21873/invivo.11587
Hajishengallis, G. (2015). Periodontitis: From microbial immune subversion to systemic inflammation. Nature Reviews. Immunology, 15(1), 30–44. https://doi.org/10.1038/nri3785
Homann, N., Tillonen, J., Rintamäki, H., Salaspuro, M., Lindqvist, C., & Meurman, J. H. (2001). Poor dental status increases acetaldehyde production from ethanol in saliva: A possible link to increased oral cancer risk among heavy drinkers. Oral Oncology, 37(2), 153–158. https://doi.org/10.1016/s1368-8375(00)00076-2
Huri, C. B., Yamalik, N., Kilinc, K., Kilinc, A., Etikan, I., & Eratalay, K. (2003). Analysis of the relationship between the severity of periodontal destruction and proteoglycan metabolism of gingiva and gingival crevicular fluid. Journal of Clinical Periodontology, 30(11), 961–968. https://doi.org/10.1034/j.1600-051X.2003.00408.x
Kalapos, M. P. (2003). On the mammalian acetone metabolism: From chemistry to clinical implications. Biochimica et Biophysica Acta, 1621(2), 122–139. https://doi.org/10.1016/s0304-4165(03)00051-5
Kc, S., Wang, X. Z., & Gallagher, J. E. (2020). Diagnostic sensitivity and specificity of host-derived salivary biomarkers in periodontal disease amongst adults: Systematic review. Journal of Clinical Periodontology, 47(3), 289–308. https://doi.org/10.1111/jcpe.13218
Kim, C., & Cha, Y. N. (2014). Taurine chloramine produced from taurine under inflammation provides anti-inflammatory and cytoprotective effects. Amino Acids, 46(1), 89–100. https://doi.org/10.1007/s00726013-1545-6
Kornman, K. S., & Papapanou, P. N. (2020). Clinical application of the new classification of periodontal diseases: Ground rules, clarifications and “gray zones”. Journal of Periodontology, 91(3), 352–360. https://doi.org/10.1002/jper.19-0557
Mainnemare, A., Mégarbane, B., Soueidan, A., Daniel, A., & Chapple, I. L. (2004). Hypochlorous acid and taurine-N-monochloramine in periodontal diseases. Journal of Dental Research, 83(11), 823–831. https:// doi.org/10.1177/154405910408301101
Mandrekar, J. N. (2010). Receiver operating characteristic curve in diagnostic test assessment. Journal of Thoracic Oncology, 5(9), 1315–1316. https://doi.org/10.1097/JTO.0b013e3181ec173d
Marcinkiewicz, J., & Kontny, E. (2014). Taurine and inflammatory diseases. Amino Acids, 46(1), 7–20. https://doi.org/10.1007/s00726-012-1361-4
Miller, C. S., Foley, J. D., Bailey, A. L., Campell, C. L., Humphries, R. L., Christodoulides, N., Floriano, P. N., Simmons, G., Bhagwandin, B., Jacobson, J. W., Redding, S. W., Ebersole, J. L., & McDevitt, J. T. (2010). Current developments in salivary diagnostics. Biomarkers in Medicine, 4(1), 171–189. https://doi.org/10.2217/bmm.09.68
Na, H. S., Kim, S. Y., Han, H., Kim, H. J., Lee, J. Y., Lee, J. H., & Chung, J. (2020). Identification of potential oral microbial biomarkers for the diagnosis of periodontitis. Journal of Clinical Medicine, 9(5), 1549. https://doi.org/10.3390/jcm9051549
Qiqiang, L., Huanxin, M., & Xuejun, G. (2012). Longitudinal study of volatile fatty acids in the gingival crevicular fluid of patients with periodontitis before and after nonsurgical therapy. Journal of Periodontal Research, 47(6), 740–749. https://doi.org/10.1111/j.1600-0765.2012.01489.x
Romano, F., Meoni, G., Manavella, V., Baima, G., Tenori, L., Cacciatore, S., & Aimetti, M. (2018). Analysis of salivary phenotypes of generalized aggressive and chronic periodontitis through nuclear magnetic resonance-based metabolomics. Journal of Periodontology, 89(12), 1452–1460. https://doi.org/10.1002/jper.18-0097
Rzeznik, M., Triba, M. N., Levy, P., Jungo, S., Botosoa, E., Duchemann, B., le Moyec, L., Bernaudin, J. F., Savarin, P., & Guez, D. (2017). Identification of a discriminative metabolomic fingerprint of potential clinical relevance in saliva of patients with periodontitis using 1H nuclear magnetic resonance (NMR) spectroscopy. PLoS One, 12(8), e0182767. https://doi.org/10.1371/journal.pone.0182767
Takahashi, N. (2015). Oral microbiome metabolism: From “who are they?” to “what are they doing?”. Journal of Dental Research, 94(12), 1628–1637. https://doi.org/10.1177/0022034515606045
Takeda, I., Stretch, C., Barnaby, P., Bhatnager, K., Rankin, K., Fu, H., Weljie, A., Jha, N., & Slupsky, C. (2009). Understanding the human salivary metabolome. NMR in Biomedicine, 22(6), 577–584. https://doi. org/10.1002/nbm.1369
Tonetti, M. S., Greenwell, H., & Kornman, K. S. (2018). Staging and grading of periodontitis: Framework and proposal of a new classification and case definition. Journal of Clinical Periodontology, 45(Suppl. 20), S149–S161. https://doi.org/10.1111/jcpe.12945
Waddington, R. J., Langley, M. S., Guida, L., Iuorio, G., Labella, R., Embery, G., & Caruso, F. (1996). Relationship of sulphated glycosaminoglycans in human gingival crevicular fluid with active periodontal disease. Journal of Periodontal Research, 31(3), 168–170. https://doi.
Walsh, M. C., Brennan, L., Malthouse, J. P., Roche, H. M., & Gibney, M. J. (2006). Effect of acute dietary standardization on the urinary, plasma, and salivary metabolomic profiles of healthy humans. The American Journal of Clinical Nutrition, 84(3), 531–539. https://doi.org/10.1093/ ajcn/84.3.5