ArticlePDF Available

Salivary Biomarkers for Dental Caries Detection and Personalized Monitoring

Authors:
  • Hahn-Schickard-Gesellschaft für angewandte Forschung e.V.

Abstract and Figures

This study investigated the potential of salivary bacterial and protein markers for evaluating the disease status in healthy individuals or patients with gingivitis or caries. Saliva samples from caries- and gingivitis-free individuals (n = 18), patients with gingivitis (n = 17), or patients with deep caries lesions (n = 38) were collected and analyzed for 44 candidate biomarkers (cytokines, chemokines, growth factors, matrix metalloproteinases, a metallopeptidase inhibitor, proteolytic enzymes, and selected oral bacteria). The resulting data were subjected to principal component analysis and used as a training set for random forest (RF) modeling. This computational analysis revealed four biomarkers (IL-4, IL-13, IL-2-RA, and eotaxin/CCL11) to be of high importance for the correct depiction of caries in 37 of 38 patients. The RF model was then used to classify 10 subjects (five caries-/gingivitis-free and five with caries), who were followed over a period of six months. The results were compared to the clinical assessments of dental specialists, revealing a high correlation between the RF prediction and the clinical classification. Due to the superior sensitivity of the RF model, there was a divergence in the prediction of two caries and four caries-/gingivitis-free subjects. These findings suggest IL-4, IL-13, IL-2-RA, and eotaxin/CCL11 as potential salivary biomarkers for identifying noninvasive caries. Furthermore, we suggest a potential association between JAK/STAT signaling and dental caries onset and progression.
Content may be subject to copyright.
Journal of
Personalized
Medicine
Article
Salivary Biomarkers for Dental Caries Detection and
Personalized Monitoring
Pune N. Paqué1, *,† , Christopher Herz 2 ,† , Daniel B. Wiedemeier 3, Konstantinos Mitsakakis 4,5 ,
Thomas Attin 1, Kai Bao 6, Georgios N. Belibasakis 6, John P. Hays 7, Joël S. Jenzer 1, Wendy E. Kaman 7,8,
Michal Karpíšek 9, Philipp Körner 1, Johannes R. Peham 2, Patrick R. Schmidlin 1, Thomas Thurnheer 1,
Florian J. Wegehaupt 1and Nagihan Bostanci 6


Citation: Paqué, P.N.; Herz, C.;
Wiedemeier, D.B.; Mitsakakis, K.;
Attin, T.; Bao, K.; Belibasakis, G.N.;
Hays, J.P.; Jenzer, J.S.; Kaman, W.E.;
et al. Salivary Biomarkers for Dental
Caries Detection and Personalized
Monitoring. J. Pers. Med. 2021,11,
235. https://doi.org/10.3390/
jpm11030235
Academic Editor: Gaetano Isola
Received: 12 February 2021
Accepted: 19 March 2021
Published: 23 March 2021
Publisher’s Note: MDPI stays neutral
with regard to jurisdictional claims in
published maps and institutional affil-
iations.
Copyright: © 2021 by the authors.
Licensee MDPI, Basel, Switzerland.
This article is an open access article
distributed under the terms and
conditions of the Creative Commons
Attribution (CC BY) license (https://
creativecommons.org/licenses/by/
4.0/).
1Clinic of Conservative and Preventive Dentistry, Center of Dental Medicine, University of Zurich,
Plattenstrasse 11, 8032 Zurich, Switzerland; thomas.attin@zzm.uzh.ch (T.A.); joel.jenzer@icloud.com (J.S.J.);
philipp.koerner@zzm.uzh.ch (P.K.); patrick.schmidlin@zzm.uzh.ch (P.R.S.);
thomas.thurnheer@zzm.uzh.ch (T.T.); florian.wegehaupt@zzm.uzh.ch (F.J.W.)
2Austrian Institute of Technology, Molecular Diagnostics, Giefinggasse 4, 1210 Wien, Austria;
christopher_herz@pall.com (C.H.); johannes.peham@ait.ac.at (J.R.P.)
3Statistical Services, Center of Dental Medicine, University of Zurich, Plattenstrasse 11,
8032 Zurich, Switzerland; daniel.wiedemeier@zzm.uzh.ch
4Hahn-Schickard, Georges-Koehler-Allee 103, 79110 Freiburg, Germany;
konstantinos.mitsakakis@hahn-schickard.de
5Laboratory for MEMS Applications, IMTEK-Department of Microsystems Engineering,
University of Freiburg, Georges-Koehler-Allee 103, 79110 Freiburg, Germany
6Department of Dental Medicine, Division of Oral Diseases, Karolinska Institutet, 141 04 Huddinge, Sweden;
kai.bao@ki.se (K.B.); george.belibasakis@ki.se (G.N.B.); nagihan.bostanci@ki.se (N.B.)
7Department of Medical Microbiology and Infectious Diseases, Erasmus University Medical Centre
Rotterdam (Erasmus MC), 3015 GD Rotterdam, The Netherlands; j.hays@erasmusmc.nl (J.P.H.);
w.e.kaman@acta.nl (W.E.K.)
8Academic Centre for Dentistry Amsterdam (ACTA), Department of Oral Biochemistry, Free University of
Amsterdam and University of Amsterdam, 1081 LA Amsterdam, The Netherlands
9BioVendor-LaboratorníMedicína, a.s., Research and Diagnostic Products Division, Immunoassays,
Clinical Validation & BioVendor Analytical Testing Service, Karasek 1767/1, 62100 Brno, Czech Republic;
karpisek@biovendor.com
*Correspondence: punenina.paque@zzm.uzh.ch; Tel.: +41-(0)4-4634-3988
These authors contributed equally to this work.
Abstract:
This study investigated the potential of salivary bacterial and protein markers for evaluating
the disease status in healthy individuals or patients with gingivitis or caries. Saliva samples from
caries- and gingivitis-free individuals (n= 18), patients with gingivitis (n= 17), or patients with
deep caries lesions (n= 38) were collected and analyzed for 44 candidate biomarkers (cytokines,
chemokines, growth factors, matrix metalloproteinases, a metallopeptidase inhibitor, proteolytic
enzymes, and selected oral bacteria). The resulting data were subjected to principal component
analysis and used as a training set for random forest (RF) modeling. This computational analysis
revealed four biomarkers (IL-4, IL-13, IL-2-RA, and eotaxin/CCL11) to be of high importance for the
correct depiction of caries in 37 of 38 patients. The RF model was then used to classify 10 subjects
(five caries-/gingivitis-free and five with caries), who were followed over a period of six months. The
results were compared to the clinical assessments of dental specialists, revealing a high correlation
between the RF prediction and the clinical classification. Due to the superior sensitivity of the RF
model, there was a divergence in the prediction of two caries and four caries-/gingivitis-free subjects.
These findings suggest IL-4, IL-13, IL-2-RA, and eotaxin/CCL11 as potential salivary biomarkers for
identifying noninvasive caries. Furthermore, we suggest a potential association between JAK/STAT
signaling and dental caries onset and progression.
Keywords:
diagnostics; interleukins; screening; personalized monitoring; saliva; biomarkers; caries;
JAK; STAT
J. Pers. Med. 2021,11, 235. https://doi.org/10.3390/jpm11030235 https://www.mdpi.com/journal/jpm
J. Pers. Med. 2021,11, 235 2 of 18
1. Introduction
The primary goal in dental medicine is the maintenance of oral health and the pre-
vention of oral diseases, such as caries or periodontitis [
1
]. Those diseases are based on
oral biofilms, which can develop on all the surfaces within the oral cavity [
2
,
3
]. A shift
in the composition of the oral microbiome paves the way for opportunistic pathogens
to induce disease outbreak and progression [
4
6
]. The development of caries lesions is
driven by microorganisms, dietary habits (the frequency of carbohydrate intake, pH and
stickiness of food debris), and host factors (salivary flow rates, immune responses, genetic
predispositions [
7
] and hygiene measures) [
8
10
]. Microorganisms produce strong organic
acids during their carbohydrate metabolism and induce mineral loss in the tooth substance
leading to tooth decay [
11
]. Progressing caries lesions cause an inflammation of the pulp,
which leads to pulpitis and periapical periodontitis. Some bacteria, such as Streptococci and
Lactobacilli, are particularly associated with the development of these caries lesions [
2
,
12
].
Persisting biofilms around the teeth on the gingival sulcus can also induce gingivitis [
13
,
14
].
This periodontal disease induces inflammatory responses in the host, causing the marginal
swelling of the gingiva, an elevated exudation of gingival crevicular fluid (GCF) and a suc-
cessive increase in local pocket depths [
15
]. Gingivitis- and periodontitis-inducing biofilms
are mainly associated with Gram-negative, anaerobic bacteria, while caries-associated bac-
teria are predominantly related with Gram-positive carbohydrate-fermenting bacteria [
15
].
However, the composition of the oral or tooth-specific microbiome may not be exclusively
associated with the maintenance of oral health or disease [
16
]. It has been shown that
patients who abstain from oral hygiene procedures subsequently develop different re-
sponses to their accumulated plaque [
13
,
17
,
18
]. These different clinical responses enabled
a classification into “periodontal-resistant” and “periodontal-insufficient” or ”high respon-
der” and “low responder”, respectively [
18
]. Hence, the host’s response to pathogenic
stimuli also accounts for the maintenance of oral health or development of disease. As such,
minor, pathogenic changes within the oral cavity can be diagnosed at an early stage, by the
molecular analysis of host immune markers. Shifts in the biomarker composition within
the GCF have been previously described in gingivitis patients [
19
21
]. For caries patients,
the close proximity of odontoblasts and the dental pulp to the lesions causes host immune
responses that lead to the production and release of several cytokines, chemokines, and
growth factors [
22
]. Both the gingivitis and the caries inflammatory responses take place in
the presence of saliva throughout the oral cavity. Thereby, this oral fluid acts as a collection
matrix for released immune defense molecules and microorganisms in both diseases.
Most studies currently focus on detecting potential biofluid targets for periodontitis
patients by using analytical procedures for biofluids such as saliva or gingival crevicular fluid.
The current consensus among dental experts is leaning towards combined and multitargeted
strategies for the identification of high-/low-risk patients, via the assessment of responses to
therapy and the prediction of periodontal stability. This may be achieved by validated combi-
nations of host-derived salivary biomarkers and key pathogens [
23
26
]. Common biomarker
panels include cytokines (such as the interleukins IL-6 and IL-1
β
), matrix metalloproteinases
(MMP-8 and MMP-9), a metallopeptidase inhibitor
(TIMP-1)
and the presence of periodontal
pathogens (e.g., Porphyromonas gingivalis,Treponema denticola,Tanerella forsythia, Fusobacterium
nucleatum,Campylobacter rectus, and Prevotella intermedia) [2733].
Knowledge of dental pulp immune response markers is usually acquired by the direct
assessment of pulp tissue and, less commonly, by the analysis of pulpal blood, peripheral
blood serum, GCF, dentinal fluid, or extracellular pulpal fluid [
22
,
34
]. Unfortunately,
relatively little is known about the combination of salivary biomarkers required to identify
caries patients or patients prone to caries [
35
,
36
]. In fact, only a few studies have indicated
differences in the salivary protein composition between healthy and caries patients [
37
39
],
with the salivary parameters investigated including basic proline-rich peptides [
37
], the
total protein load [
38
], and the total protein composition by molecular weight pattern [
39
].
Molecular analyses for salivary caries signatures remain comparatively unexplored.
J. Pers. Med. 2021,11, 235 3 of 18
The direct contact of saliva with oral lesions, as well as saliva’s simple availability,
highlights its potential use as a diagnostic medium for the detection of oral diseases [
40
,
41
].
Furthermore, due to the above-mentioned multifactorial etiology of caries and associated
challenges in the development of prevention and treatment strategies, there is a need for
rapid personalized diagnostics [
42
,
43
]. In fact, the discovery of suitable salivary biomarkers
might enable the early identification of high-risk patients prone to developing gingivi-
tis, whilst allowing the early detection of caries lesion development. Furthermore, such
biomarkers could be used to assess the status of orally healthy individuals, including in
patient monitoring, over time.
Therefore, the current study set out to analyze the saliva of 73 subjects, who were
grouped based on clinical diagnosis into healthy individuals, patients with gingivitis and
patients with untreated deep caries lesions. Central to the study was an exploratory multi-
method approach aiming to identify suitable biomarkers that could predict dental caries.
A secondary aim was to detect gingivitis patients within the same cohort. Additionally,
10 subjects (five caries-/gingivitis-free and five with caries) were followed over a period
of six months, and saliva was collected for biomarker analysis. It was hypothesized that
a combination of protein and microbial candidate markers could enable a differentiation
between all groups and be further validated with the 10 follow-up subjects.
2. Materials and Methods
2.1. Study Population and Study Design
Individuals who came for a check-up or dental treatment to the Center of Dental
Medicine, Zurich, and who fulfilled the inclusion/exclusion criteria were recruited for
the study (approved by the local Swiss ethics committee with the BASEC-no. 2016-00435,
date of approval: 09.01.2016). A total of 120 subjects (Figure 1) were asked to participate
in a previous study (minimum age of 18 years; mean, 35.8 years; 57 females) and donate
saliva for a microbial analysis and identification of diseases with a qPCR (quantitative
polymerase chain reaction) assay [
12
]. Out of this cohort, stored (
80
C, not longer than
six months) saliva was used from healthy subjects (n= 29; mean age, 31.8; 18 females),
patients with gingivitis (n= 22; mean age, 35.4; 13 females), and patients with open
caries lesions (n= 69; mean age, 37.6; 26 females). Complete data for all the applied
methods (
44 biomarkers
) were available for 73 of the initially collected 120 samples and
were included in the computational prediction analysis: healthy patients (n= 18), patients
with gingivitis (n= 17), and open caries lesions (n= 38); see Section 2.7. The participating
subjects visited the Center of Dental Medicine twice. The first appointment (30–60 min)
included a dental examination by the treating dentist. Eligible subjects who signed written
informed consent visited the Center for a second appointment (within two weeks), during
which saliva collection took place.
The general health status and potential in- and exclusion criteria of the subjects were
assessed by a dentist during the first visit. Systemically healthy subjects were asked to
participate in this study with a selected oral health status (general health was indicated by
the exclusion of patients with diabetes; heart disease; infections, such as tuberculosis, hep-
atitis, sexually transmitted diseases, and HIV/AIDS; tumor diseases; and gastric digestive
disorders causing vomiting). Subjects were excluded if they had already participated in an-
other clinical trial within the last three months prior to the collection of saliva, were heavy
smokers taking > 10 cigarettes per day (e-cigarette usage was not separately surveyed),
or were pregnant or lactating women. The clinical examination allowed a grouping of
subjects into orally healthy (no caries lesions, and periodontal screening index (PSI) scores
of 0 in
4
sextants and
2 in
2 sextants), gingivitis (no caries lesions, and PSI scores of 1
in
4
sextants and
2 in
2 sextants), or caries patients (
2 open dentinal caries lesions
and PSI scores of 2 in 6 sextants).
Patient enrollment and saliva collection took place from September 2018 to April
2019 and were approved by the local Swiss ethics committee (BASEC-no. 2016-00435).
Each subject donated saliva once (baseline). For further investigations, 10 subjects (five
J. Pers. Med. 2021,11, 235 4 of 18
caries-/gingivitis-free and five with caries) of the cohort were asked to repeat the saliva
donation four, five, and six months after baseline. The donated saliva of each participating
subject was anonymized during collection, and access to patient data was only available to
authorized study personnel. All the saliva samples were screened for 44 candidate biomark-
ers (cytokines, chemokines, growth factors, matrix metalloproteinases, a metallopeptidase
inhibitor, proteolytic enzymes, and selected oral bacteria). These biomarkers were selected
as a multitargeted approach to addressing different immune responses, which were, in
part, already described as correlating with oral diseases [2733].
J. Pers. Med. 2021, 11, x FOR PEER REVIEW 4 of 19
Figure 1. Flow-chart showing the study design, study population, and applied saliva analyses. Biomarkers of the Cytokine
30-Plex Human Panel are color-coded for cytokines (green), chemokines (yellow), and growth factors (blue). * Eo-
taxin/CCL11.
Patient enrollment and saliva collection took place from September 2018 to April 2019
and were approved by the local Swiss ethics committee (BASEC-no. 2016-00435). Each
subject donated saliva once (baseline). For further investigations, 10 subjects (five caries-
/gingivitis-free and five with caries) of the cohort were asked to repeat the saliva donation
four, five, and six months after baseline. The donated saliva of each participating subject
was anonymized during collection, and access to patient data was only available to au-
thorized study personnel. All the saliva samples were screened for 44 candidate bi-
omarkers (cytokines, chemokines, growth factors, matrix metalloproteinases, a metallo-
peptidase inhibitor, proteolytic enzymes, and selected oral bacteria). These biomarkers
were selected as a multitargeted approach to addressing different immune responses,
which were, in part, already described as correlating with oral diseases [27–33].
Figure 1.
Flow-chart showing the study design, study population, and applied saliva analyses. Biomarkers of the
Cytokine 30-Plex Human Panel are color-coded for cytokines (green), chemokines (yellow), and growth factors (blue).
* Eotaxin/CCL11.
2.2. Saliva Sampling
2.2.1. Baseline Collection of Saliva
As described in a previous study [
12
], a strict saliva donation protocol was followed.
Briefly, the participants were asked not to eat, drink sugary drinks, or perform any oral hy-
giene measures the night before the saliva donation. However, water intake was permitted
J. Pers. Med. 2021,11, 235 5 of 18
at all times. A standardized saliva donation was timed between 8.00 and 10.00 am [44,45]
and performed to obtain unstimulated whole saliva [
46
]. A video was provided to in-
struct the subjects on the methodology for standardized saliva donation. A collection
time of 15 min was scheduled, and the saliva was collected in a test tube. Tubes con-
taining less than 1.8 mL of saliva were discarded. Tubes with sufficient volumes were
then vortexed, aliquoted in DNA low-bind tubes or protein low-bind tubes (Eppendorf,
Wesseling-Berzdorf, Germany) depending on the following assay, and stored at 80 C.
2.2.2. Repeated Measurements of Saliva in Selected Subjects
Ten representative subjects of the cohort (five caries-/gingivitis-free and five with
caries) were followed over a period of six months (Figure 1). Analogous to timepoint
0 (baseline), the patients received three further oral checkups and were grouped into their
respective clinical groups, i.e., “healthy”, “gingivitis”, or “caries”. During the following
observation time, three successive saliva collections took place, at four, five, and six months
after the baseline collection. The patients were asked at each saliva collection appointment
to answer questions regarding the dental treatments performed, such as the treatment
of caries lesions, the extraction of teeth, and whether a professional tooth cleaning had
taken place. Additionally, the patients were asked about antibiotic intake, changes in
their nutrition regimes, and their self-care plaque control. The oral health statuses of the
patients were assessed during each saliva collection appointment by dental specialists.
Saliva samples were collected following the same protocol as used for the baseline saliva
collection, coded, and stored for further analysis of the 44 markers as previously described.
The resulting dataset consisting of the repeated measurements of 10 additional individuals
was later used as an independent dataset to check the statistical models and is therefore
also referred to as the “test dataset”.
2.3. 30-Plex Analysis of Biomarkers in Saliva
For each sample, the aliquot with the highest volume was thawed from
80
C
overnight at 6
C on an orbital shaker at 180 rpm to ensure homogeneity. It was then
centrifuged at 21,000 rcf for 3 min at 4
C. A total of 25
µ
L of the supernatant was used for
the analysis with the Cytokine 30-Plex Human Panel (Cat. # LHC6003M, by Thermo Fisher
Scientific, Supplementary Table S1). The manufacturer’s recommendations were followed
throughout the analysis. Cytokine capturing was performed by incubating overnight. The
biotinylated antibody step was extended to 1 h, and the R-phycoerythrin (R-PE) coupling,
to 2 h. The instrumental setup was established according to the 30-Plex Human Panel
protocol with a target bead count of 50. Data acquisition was carried out on a Luminex
200 system (Invitrogen, 4088 Commercial Ave, Northbrook, IL, USA) and its xPONENT
software (Version 3.1, Invitrogen, Waltham, MA, USA) package. The plates were measured
twice, on both low and high Photomultiplier Tube (PMT) settings.
2.4. MMP-8, MMP-9, and TIMP-1 Measurements
The MMP-8, MMP-9, and TIMP-1 concentrations were determined by using commer-
cial ELISA kits (BioVendor-Laboratornímedicína a.s., Brno, Czech Republic), according
to the manufacturer’s instructions [
47
,
48
]. The calibration range was adapted for the de-
termination of MMP-8 in saliva samples. Saliva supernatant from each patient sample
was centrifuged at 4
C for 10 min at 10,000
×
gand then diluted 20, 40, or 250 times,
respectively, with an appropriate dilution buffer, and used for analysis.
2.5. Protease Analysis
Total salivary proteolytic activity was measured as described previously [
49
]. In brief, 49
µ
L
of saliva was incubated with 1
µ
L of 800
µ
M PEK-054 ([FITC]-NleKKKKVLPIQLNAATDK-
[KDbc]), a substrate used to assess total protease activity. A trypsin solution from bovine
pancreas (500 U, Sigma-Aldrich Chemie B.V., Zwijndrecht, The Netherlands) was used as a
standard. The reactions were incubated at 37
C for 60 min, and the increase in fluorescence
J. Pers. Med. 2021,11, 235 6 of 18
was determined using a fluorescence microplate reader (FLUOstar Galaxy, BMG Laboratories)
at an excitation wavelength of 485 nm and an emission wavelength of 530 nm. All the saliva
samples were analyzed in duplicate.
2.6. qPCR
2.6.1. DNA Extraction for qPCR
DNA was extracted using the GenElute
TM
Bacterial Genomic DNA Kit (Sigma-Aldrich,
St. Louis, MO, USA) and the protocol for Gram-positive bacterial preparation and Strepto-
coccus species (with the addition of 250 units/mL mutanolysin) with a prolonged lysis step.
For each patient, 920
µ
L of saliva was spun down at 18,000 rcf for 3 min. The remaining
pellet was resuspended in an enzyme solution consisting of lysozyme and mutanolysin and
incubated for 1 h at 37
C on a Thermomixer (1400 rpm, Eppendorf, Wesseling-Berzdorf,
Germany). RNase treatment was conducted according to the manufacturer’s recommen-
dations. The incubation time during proteinase K treatment was prolonged to 30 min at
55
C at 1400 rpm. The DNA was eluted in 135
µ
L of 10 mM TrisHCl (pH 8.8) and stored at
25 C.
2.6.2. Oligonucleotide Design and Primer and Probe Specificity
Novel primer and probe sets were designed for the bacterial strains P. gingivalis,
T. forsythia, T. denticola, F. nucleatum, C. rectus, P. intermedia, A. actinomycetemcomitans,
S. mutans, S. sobrinus and orally associated Lactobacilli [
12
]. Target sequences were
extracted out of the “Nucleotide” database [
50
] provided by the National Center for Biotech-
nology Information (NCBI) and the “Human Oral Microbiome Database” (HOMD), courtesy
of the Forsyth Institute in Cambridge, MA, USA [
5
]. The primer and probe sets targeted the
16S ribosomal RNA gene except for A. actinomycetemcomitans and S. sobrinus. The limited
availability and heterogeneity of the A. actinomycetemcomitans 16S rRNA sequences in the
NCBI Nucleotide database meant that the virulence factor and toxin gene LtxA was targeted
instead. Streptococci display a high sequence similarity within their genera; hence, the 23S
rRNA (uracil-5-)-methyltransferase RumA gene was selected to increase the specificity of
detection. The oligonucleotide design was exclusively performed using Primer3 [
51
]. The
oligonucleotide coverage, as well as specificity, was first confirmed in silico using Primer
Blast in combination with the “not redundant/nucleotide (nr/nt)”, “chromosomes of all
organism”, and “HOMD 16S rRNA RefSeq” databases. The results were validated in the lab
using bacterial genomic reference DNA provided by the Leibniz Institute DSMZ, German
Collection of Microorganisms and Cell Cultures GmbH (https://www.dsmz.de/), and the
ATCC, American Type Culture Collection (www.lgcstandards-atcc.org) [12].
2.6.3. Duplex qPCR
qPCR was performed using a dual-color format utilizing a custom “TaqMan Lyophilized
1-Step qPCR MasterMix” on a Roche Light Cycler 480II. Each reaction contained extracted
patient material or reference genomic DNA and an internal amplification control, which
consisted of 0.01 ng of genomic DNA of Serinicoccus marinus. Probes targeting orally
associated bacteria were labelled with 6-carboxyfluorescein (6-FAM), whereas the internal
control probe used Roche’s proprietary LightCycler Red 610. A two-step cycling protocol
was applied, starting with an initial activation for 2 min at 95
C followed by 40 cycles
alternating between 95
C for 3 s and 60
C for 30 s. First, external standard curves were
generated in three consecutive qPCR runs for each oral taxon using the bacterial genomic
reference DNA highlighted in Section 2.6.2. The standards included six concentrations
in 10-fold increments from 10 ng to 0.1 pg [
12
]. The values for each standard were then
converted into genome equivalents by extracting the genome size of each strain from the
NCBI Assembly [
52
], multiplying it by the average molecular weight of a base pair, and cor-
relating it to the six standard concentrations. Subsequently, the extracted patient DNA was
subjected to qPCR, and the corresponding Cq values were directly quantified and converted
into genome equivalents on the basis of the previously established standard curves.
J. Pers. Med. 2021,11, 235 7 of 18
2.7. Statistics and Computational Analysis
The data of a saliva sample were only used if all the molecular methods were ap-
plied on that sample, resulting in the measurement of 44 markers in 73 samples. All the
measurements are expressed as genome equivalents, absorbance units (AU), pg/mL, or
ng/mL (depending on the biomarker) and compiled in a spreadsheet. Only data above
the detection limit (depending on each biomarker) were used directly (Supplementary
Table S2). Measurements below the detection limit were imputed using the half-minimum
method [53].
The baseline data (“training dataset”) were used for principal component analysis
(PCA) on the correlation matrix and for training the random forest (RF). Moreover, this
dataset was also the basis for comparing the marker levels of IL-4, IL-13, IL-2-RA, and
eotaxin/CCL11 between healthy individuals, patients with gingivitis, and patients with
untreated deep caries lesions using nonparametric one-way ANOVAs (Kruskal–Wallis
rank sum tests) followed by pairwise comparisons according to Conover. The additional
data from recalling 10 subjects (“test dataset”) were prepared in the same way, but were
only used to check RF predictions. RF was chosen over other algorithms because it offers
powerful prediction properties in combination with fairly interpretable model character-
istics and fits [
54
,
55
]. All the computational analyses were performed with the statistical
software R [
56
], including the packages ggplot2 [
57
], PMCMRplus [
58
], randomForest [
59
],
and FactoMineR [60].
3. Results
The saliva samples of 120 subjects were collected and analyzed with different molec-
ular methods, screening for 19 cytokines, 7 chemokines, 4 growth factors, 2 metallo-
proteinases, 1 metallopeptidase inhibitor, 1 protease, and 10 orally associated bacteria
(Figure 1).
Data were only subjected to further statistical modeling if results for all the
assays (44 prominent protein and DNA markers) were available. Altogether, this accounted
for 73 saliva samples, of which 18 were grouped as healthy, 17 as gingivitis, and 38 as
caries patients.
3.1. Frequency of Detected Biomarkers
Data were gathered at various time points by project partners throughout this study.
An initial visualization enabled the assessment of data quality as shown in the detection
limit matrix (Figure 2). It can be seen that data for A. actinomycetemcomitans, S. sobrinus,
and oral Lactobacilli were below the detection limit in the majority of the subjects. Fur-
thermore, there were no data for all the IL-1-RA and IL-5 measurements. The remaining
assays, however, showed a very robust performance (data above the detection limit; see
Supplementary Table S2).
3.2. Principal Component Analysis (PCA)
The first principal component encompassed 36% of the total variance in the data, while
the second principal component absorbed 8.4% of the total variance (Figure 3). There was a
profound overlap between the healthy and gingivitis groups, while the caries group could
generally be better differentiated. This suggested that the groups exhibited distinctive
marker profiles.
J. Pers. Med. 2021,11, 235 8 of 18
J. Pers. Med. 2021, 11, x FOR PEER REVIEW 8 of 19
Figure 2. Detection matrix of all biomarkers (x-axis) in the analyzed 73 subjects (y-axis). Color
codes show data measured above (blue) or below (red) the detection limit. Numbers on the left
correspond to the numbers of patients with the same pattern of biomarkers above/below the detec-
tion limit. Numbers on the right represent the numbers of biomarkers below the detection limit
per line.
3.2. Principal Component Analysis (PCA)
The first principal component encompassed 36% of the total variance in the data,
while the second principal component absorbed 8.4% of the total variance (Figure 3).
There was a profound overlap between the healthy and gingivitis groups, while the caries
group could generally be better differentiated. This suggested that the groups exhibited
distinctive marker profiles.
Figure 2.
Detection matrix of all biomarkers (x-axis) in the analyzed 73 subjects (y-axis). Color codes
show data measured above (blue) or below (red) the detection limit. Numbers on the left correspond
to the numbers of patients with the same pattern of biomarkers above/below the detection limit.
Numbers on the right represent the numbers of biomarkers below the detection limit per line.
J. Pers. Med. 2021, 11, x FOR PEER REVIEW 9 of 19
Figure 3. Results obtained from a principal component analysis (PCA) with the first dimension on the x-axis (36%) and
the second dimension on the y-axis (8.4%). The groups are separated by color and shape (caries = red square; gingivitis =
green triangle; caries = red circle). Three bigger shapes are located at the center of gravity for each group.
3.3. Biomarkers for Discriminating the Oral Health Status
Following the PCA, another multidimensional approach (a random forest supervised
learning model) was applied to the training dataset in order to extract the strongest dis-
criminators that were potentially capable of correctly assigning patients into the three clin-
ical study groups. Figure 4 illustrates the strongest discriminators plotted according to the
mean decrease in accuracy: IL-4, IL-13, and IL-2-RA, followed by eotaxin/CCL11. These
were the four classifiers that enabled the best distinction of caries, healthy, and gingivitis
patients considering all the measured markers.
Figure 3.
Results obtained from a principal component analysis (PCA) with the first dimension on the x-axis (36%) and the
second dimension on the y-axis (8.4%). The groups are separated by color and shape (caries = red square; gingivitis = green
triangle; caries = red circle). Three bigger shapes are located at the center of gravity for each group.
J. Pers. Med. 2021,11, 235 9 of 18
3.3. Biomarkers for Discriminating the Oral Health Status
Following the PCA, another multidimensional approach (a random forest supervised
learning model) was applied to the training dataset in order to extract the strongest discrim-
inators that were potentially capable of correctly assigning patients into the three clinical
study groups. Figure 4illustrates the strongest discriminators plotted according to the
mean decrease in accuracy: IL-4, IL-13, and IL-2-RA, followed by eotaxin/CCL11. These
were the four classifiers that enabled the best distinction of caries, healthy, and gingivitis
patients considering all the measured markers.
J. Pers. Med. 2021, 11, x FOR PEER REVIEW 10 of 19
Figure 4. Random forest results plotted according to mean decrease in accuracy, show the strong-
est discriminators, namely, IL-4, IL-13, IL-2-RA, and eotaxin/CCL11.
Interestingly, only four markers were needed to generate the best discrimination be-
tween the different oral health groups, i.e., IL-4, IL-13, and IL-2-RA as well as the chemo-
kine eotaxin/CCL11 (Figure 4). Oral bacteria did not play an important role. No statisti-
cally significant differences were found between healthy and gingivitis patients regarding
the tested biomarkers (Table 1). Nevertheless, there was a clear statistically significant dif-
ference between the caries and the other two clinical groups (Figure 5, Table 1). Elevated
biomarker levels could be observed for all four biomarkers in caries, when compared to
those in the healthy and gingivitis groups. The mean data ± standard deviation was cal-
culated for all the biomarkers. The biomarker levels of the patients with deep caries lesions
were 44.9 pg/mL ± 6.8 pg/mL for IL-4, 4.6 pg/mL ± 2.6 pg/mL for IL-13, 158.9 pg/mL ± 44.4
pg/mL for IL-2-RA, and 1.7 pg/mL ± 0.9 pg/mL for eotaxin/CCL11. The biomarker IL-13
was below the detection limit in the healthy and gingivitis groups.
Figure 4.
Random forest results plotted according to mean decrease in accuracy, show the strongest
discriminators, namely, IL-4, IL-13, IL-2-RA, and eotaxin/CCL11.
Interestingly, only four markers were needed to generate the best discrimination
between the different oral health groups, i.e., IL-4, IL-13, and IL-2-RA as well as the
chemokine eotaxin/CCL11 (Figure 4). Oral bacteria did not play an important role. No
statistically significant differences were found between healthy and gingivitis patients
regarding the tested biomarkers (Table 1). Nevertheless, there was a clear statistically sig-
nificant difference between the caries and the other two clinical groups (Figure 5,
Table 1).
Elevated biomarker levels could be observed for all four biomarkers in caries, when com-
pared to those in the healthy and gingivitis groups. The mean data
±
standard deviation
was calculated for all the biomarkers. The biomarker levels of the patients with deep
caries lesions were 44.9 pg/mL
±
6.8 pg/mL for IL-4, 4.6 pg/mL
±
2.6 pg/mL for IL-13,
J. Pers. Med. 2021,11, 235 10 of 18
158.9 pg/mL ±44.4 pg/mL
for IL-2-RA, and 1.7 pg/mL
±
0.9 pg/mL for eotaxin/CCL11.
The biomarker IL-13 was below the detection limit in the healthy and gingivitis groups.
Table 1.
p-values of the four biomarkers IL-4, IL-13, IL-2-RA, and eotaxin/CCL11 when comparing
patients with caries, patients with gingivitis, and healthy individuals.
Biomarkers Caries/Gingivitis Caries/Healthy Gingivitis/Healthy
IL-4 1.5 ×1015 4.1 ×1013 0.17
IL-13 4.0 ×1013 3.1 ×1012 0.52
IL-2-RA 3.3 ×1061.0 ×1040.35
Eotaxin/CCL11 8.1 ×1054.4 ×1040.56
J. Pers. Med. 2021, 11, x FOR PEER REVIEW 11 of 19
Figure 5. Boxplots of the four strongest group classifiers (IL-4, IL-13, IL-2-RA, and eotaxin/CCL11) based on random forest
classification, with median values and interquartile ranges for each group shown (healthy individuals, patients with gin-
givitis, and those with caries). p-values were derived from Kruskal–Wallis tests followed by post hoc pairwise comparisons
according to Conover. ns = p > 0.05; *** p 0.001 (see Table 1 for specific p-values).
Table 1. p-values of the four biomarkers IL-4, IL-13, IL-2-RA, and eotaxin/CCL11 when comparing
patients with caries, patients with gingivitis, and healthy individuals.
Biomarkers Caries/Gingivitis Caries/Healthy Gingivitis/Healthy
IL-4 1.5 × 10
15
4.1 × 10
13
0.17
IL-13 4.0 × 10
13
3.1 × 10
12
0.52
IL-2-RA 3.3 × 10
6
1.0 × 10
4
0.35
Eotaxin/CCL11 8.1 × 10
5
4.4 × 10
4
0.56
3.4. PCA Loading Plot
PCA loading plots were generated to visualize potential correlations between the
tested biomarkers and clusters of samples. The directions and angles of the plots showed
a positive correlation between IL-2-RA and IL-13, followed by IL-4 (Figure 6). Another
interesting observation is that periodontitis and caries-associated bacteria are not likely to
be correlated, although, as expected, bacterial protease activity appeared to correlate bet-
ter with the periodontitis-associated strains. The only caries-associated strain showing a
positive correlation with IL-4, IL-13, and IL-2-RA was Streptococcus mutans (green).
Figure 5.
Boxplots of the four strongest group classifiers (IL-4, IL-13, IL-2-RA, and eotaxin/CCL11) based on random
forest classification, with median values and interquartile ranges for each group shown (healthy individuals, patients
with gingivitis, and those with caries). p-values were derived from Kruskal–Wallis tests followed by post hoc pairwise
comparisons according to Conover. ns = p> 0.05; *** p0.001 (see Table 1for specific p-values).
3.4. PCA Loading Plot
PCA loading plots were generated to visualize potential correlations between the
tested biomarkers and clusters of samples. The directions and angles of the plots showed
a positive correlation between IL-2-RA and IL-13, followed by IL-4 (Figure 6). Another
interesting observation is that periodontitis and caries-associated bacteria are not likely
to be correlated, although, as expected, bacterial protease activity appeared to correlate
better with the periodontitis-associated strains. The only caries-associated strain showing
a positive correlation with IL-4, IL-13, and IL-2-RA was Streptococcus mutans (green).
J. Pers. Med. 2021,11, 235 11 of 18
J. Pers. Med. 2021, 11, x FOR PEER REVIEW 12 of 19
Figure 6. PCA loading plot visualizing the correlation between the biomarkers tested and clusters of samples, grouped
based on their similarity. The red-colored arrows show the four biomarkers (ordered from top to bottom): IL-4, IL-13, IL-
2-RA, and eotaxin/CCL11. The S. mutans arrow is colored in green, in close proximity to IL-4.
An objective of this study was to utilize the measured biomarkers in order to predict
the clinical health statuses of patients (Table 1). Using the RF algorithm, the results were
then compared to the initial clinical assessments of dental specialists. The overall out-of-
bag error amounted to 30%. However, there was a pronounced contrast in classification
error between the individual groups, as can be seen in Table 2. The caries group could be
clearly differentiated from the gingivitis and the healthy groups (classification error of
2.6%), while the gingivitis and healthy groups could not be satisfactorily distinguished
from each other on the basis of the biomarkers that were investigated (classification errors
of 58.8% and 61.1%, respectively).
Table 2. Confusion matrix comparing random forest (RF) model predictions using biomarkers
(top) with the assessments by dental specialists (left).
Status Healthy Gingivitis Caries Classification Error (%)
Healthy 7 10 1 58.8
Gingivitis 8 7 2 61.1
Caries 1 0 37 2.6
Figure 6.
PCA loading plot visualizing the correlation between the biomarkers tested and clusters of samples, grouped
based on their similarity. The red-colored arrows show the four biomarkers (ordered from top to bottom): IL-4, IL-13,
IL-2-RA, and eotaxin/CCL11. The S. mutans arrow is colored in green, in close proximity to IL-4.
An objective of this study was to utilize the measured biomarkers in order to predict
the clinical health statuses of patients (Table 1). Using the RF algorithm, the results were
then compared to the initial clinical assessments of dental specialists. The overall out-of-bag
error amounted to 30%. However, there was a pronounced contrast in classification error
between the individual groups, as can be seen in Table 2. The caries group could be clearly
differentiated from the gingivitis and the healthy groups (classification error of 2.6%), while
the gingivitis and healthy groups could not be satisfactorily distinguished from each other
on the basis of the biomarkers that were investigated (classification errors of 58.8% and
61.1%, respectively).
Table 2.
Confusion matrix comparing random forest (RF) model predictions using biomarkers (top)
with the assessments by dental specialists (left).
Status Healthy Gingivitis Caries Classification Error (%)
Healthy 7 10 1 58.8
Gingivitis 8 7 2 61.1
Caries 1 0 37 2.6
J. Pers. Med. 2021,11, 235 12 of 18
3.5. Comparison of the RF Model Predictions and Clinical Assessments
RF model prediction was applied on a cohort of 10 subjects (five caries-/gingivitis-free
and five with caries) that were part of the repeated measurement group (test dataset). The
computational predictions were then compared to clinical classifications performed during
the saliva collection appointments by dental practitioners. In addition to timepoint 0 saliva
measurements (T0, baseline), three successive measurements were taken four months (T1),
five months (T2), and six months (T3) after T0. The comparison of the classifications by
the RF model and the clinical classification is visualized in Figure 7. The RF classification
of healthy subjects was identical to the clinical classification for 13 of the 20 examination
points (Figure 7, Healthy Patients). Minor divergences within the healthy group could be
observed in seven instances. The model predicted a condition of gingivitis contrary to to
the clinical specialist.
J. Pers. Med. 2021, 11, x FOR PEER REVIEW 14 of 19
Figure 7. The clinical statuses of five healthy and five caries patients were assessed and classified
by dental specialists as well as by biomarker-based RF predictions (Healthy Patient_1 to _5, and
Cariogenic Patient_1 to _5). The graphs visualize and compare the classification results based on
the RF modeling (white) and the clinical assessment by dental specialists (black) for each
timepoint (T0 = baseline, T1 = after 4 months, T2 = after 5 months, and T3 = after 6 months). The
tables below the graph show the respective biomarker concentrations in pg/mL, with N/D = values
below the detection limit.
4. Discussion
Out of 44 potential biomarkers, a total of four salivary biomarkers were found to ex-
hibit strong potential as classifiers for differentiating between healthy individuals and
caries patients. These were the interleukins IL-4 and IL-13, the interleukin receptor IL-2-
RA, and the chemokine eotaxin/CCL11. Using, mainly, these four biomarkers, caries pa-
tients could be classified into the correct group with a very high degree of certainty (clas-
sification error of 2.6%). The RF prediction was based on a training set of 73 subjects and
used for the health assessment and prediction of 10 individuals (five caries-/gingivitis-free
and five with caries), who were followed over a period of six months (test dataset). The
results suggest that the biomarker-based RF prediction of caries patients is more sensitive
than clinical assessments by dental specialists. A distinct discrimination between the
healthy and gingivitis groups was not possible in this study. Therefore, the initial hypoth-
esis was partially rejected. A clear differentiation of the caries group based on protein but
not bacterial biomarkers was enabled and validated with the 10 follow-up subjects.
With respect to IL-4 and IL-13, since 1993, it has been established that the IL-13 gene
is closely linked to the IL-4 gene on chromosome 5q 23-31 [61,62], and that there is se-
quence homology and a shared subunit responsible for signal transduction between these
two genes. It has been suggested that both IL-4 and IL-13 are potent in vitro modulators,
critical in the regulation of primarily Th2 immune responses [63]. It was shown that Th2
cytokines, such as IL-4, IL-5, and IL-13, together with eotaxin/CCL11, regulate critical as-
pects of eosinophil recruitment, allergic inflammation, and airway hyperresponsiveness
in asthma [62]. A recent study of human monocytes and macrophages confirmed that IL-
13 utilizes both the IL-4-RA–JAK2–STAT3 and IL-13-RA–Tyk2–STAT1/STAT6 signaling
cascades (JAK = Janus kinase; STAT = signal transducer and activator of transcription pro-
teins), whereas IL-4 can only use the IL-4-RA–JAK1–STAT3/STAT6 axis [64]. A different
group investigated the potential of IL-13 as an activator of JAK3–STAT6 signaling in cells
Figure 7.
The clinical statuses of five healthy and five caries patients were assessed and classified
by dental specialists as well as by biomarker-based RF predictions (Healthy Patient_1 to _5, and
Cariogenic Patient_1 to _5). The graphs visualize and compare the classification results based on
the RF modeling (white) and the clinical assessment by dental specialists (black) for each timepoint
(T0 = baseline, T1 = after 4 months, T2 = after 5 months, and T3 = after 6 months). The tables below
the graph show the respective biomarker concentrations in pg/mL, with N/D = values below the
detection limit.
J. Pers. Med. 2021,11, 235 13 of 18
All the caries patients that were followed up during the study changed group during
the repeated saliva measurements—from caries to healthy—as indicated by their dental
practitioners. The clinical examinations of these patients eventually revealed a caries-free
dentition. Changes in diet or self-care plaque control, antibiotic intake, tooth extraction,
or professional tooth cleaning were documented and are available in the Supplementary
Table S3.
Biomarker-based RF prediction for these clinically assessed orally healthy patients
generated a different conclusion for four patients (Cariogenic Patient_1 with T2 and T3
Caries, and Cariogenic Patient_4 with T2 and T3 Caries). Based on the RF model and the
salivary biomarker levels, these patients remained classified as caries patients (Figure 7).
However, overall, there was good agreement, with 16 out of 20 classifications matching
between the dental experts and the RF model.
4. Discussion
Out of 44 potential biomarkers, a total of four salivary biomarkers were found to
exhibit strong potential as classifiers for differentiating between healthy individuals and
caries patients. These were the interleukins IL-4 and IL-13, the interleukin receptor IL-2-RA,
and the chemokine eotaxin/CCL11. Using, mainly, these four biomarkers, caries patients
could be classified into the correct group with a very high degree of certainty (classification
error of 2.6%). The RF prediction was based on a training set of 73 subjects and used for
the health assessment and prediction of 10 individuals (five caries-/gingivitis-free and five
with caries), who were followed over a period of six months (test dataset). The results
suggest that the biomarker-based RF prediction of caries patients is more sensitive than
clinical assessments by dental specialists. A distinct discrimination between the healthy
and gingivitis groups was not possible in this study. Therefore, the initial hypothesis was
partially rejected. A clear differentiation of the caries group based on protein but not
bacterial biomarkers was enabled and validated with the 10 follow-up subjects.
With respect to IL-4 and IL-13, since 1993, it has been established that the IL-13 gene is
closely linked to the IL-4 gene on chromosome 5q 23-31 [
61
,
62
], and that there is sequence
homology and a shared subunit responsible for signal transduction between these two
genes. It has been suggested that both IL-4 and IL-13 are potent
in vitro
modulators,
critical in the regulation of primarily Th2 immune responses [
63
]. It was shown that Th2
cytokines, such as IL-4, IL-5, and IL-13, together with eotaxin/CCL11, regulate critical
aspects of eosinophil recruitment, allergic inflammation, and airway hyperresponsiveness
in asthma [
62
]. A recent study of human monocytes and macrophages confirmed that
IL-13 utilizes both the IL-4-RA–JAK2–STAT3 and IL-13-RA–Tyk2–STAT1/STAT6 signaling
cascades (JAK = Janus kinase; STAT = signal transducer and activator of transcription
proteins), whereas IL-4 can only use the IL-4-RA–JAK1–STAT3/STAT6 axis [
64
]. A different
group investigated the potential of IL-13 as an activator of JAK3–STAT6 signaling in
cells expressing IL-2-RG and IL-4-RA [
65
]. Although initially not stated, Thermo Fisher
Scientific confirmed that the utilized “Cytokine 30-Plex Human Panel” assay was developed
using the IL-2-RA (Accession Number: P01589) protein. A potential cross-reactivity and
unintentional measurement of IL-2-RG in addition to IL-2-RA should be further assessed.
The assumption that JAK–STAT signaling could potentially play a role in dental caries
was supported by a study on human periodontal ligament cells (HPDLCs). The study
revealed that IL-4 is essential in STAT6 activation and the release of the eosinophil-specific
chemoattractant eotaxin/CCL11 [66].
Notably, from our results, four of the 20 follow-up measurements in the caries group
differed between the clinical assessments by dentists and the computational predictions
based on the biomarkers. The clinical examinations revealed a caries-free dentition after
placing fillings on the open caries lesions of two patients (Cariogenic Patient_1 and Cari-
ogenic Patient_4 after five and six months). Interestingly, both patients initially showed
diverse caries lesions on many teeth simultaneously, meaning that caries treatment was
performed on all the open caries lesions. Once they were treated, dentists classified these
patients as healthy, according to the study classification criteria. However, molecular
J. Pers. Med. 2021,11, 235 14 of 18
analysis indicated the presence of ongoing caries-related immune responses. Persisting
biomarkers, which were detected in the saliva of these two multi-caries-treated patients,
may have been triggered due to the multiple initial caries lesions and originated from open
dentinal tubules or the GCF. Furthermore, the differences in the classifications of healthy
patients might be related to the clinical classification protocol utilized during the initial
appointments and the enrollment of the study subjects, as the dental examination included
screening for caries lesions and their validation by radiographs, which enabled a clear
classification. The healthy and gingivitis groups, however, were only characterized by the
absence of caries lesions and dental pockets over 3 mm in depth. Bleeding on probing,
which is the key parameter for classifying gingivitis patients, was only documented per
sextant using PSI scores (gingivitis grouping: no caries lesions, and PSI scores of 1 in
4 sextants
and
2 in
2 sextants). This clinical differentiation may have been improved
if other bleeding scores had been employed [67].
The low impact of bacterial population levels on the overall classification power
of the RF model can potentially be explained by several different aspects. To an extent,
periopathogens were not expected to play a role in the differentiation and classification of
caries. They were included in this study due to their potential involvement in gingivitis.
While no sole bacterial species can be singled out as the causative factor for gingivitis [
68
],
the bacterial species selected for these assays have long been known to be associated
with periodontal diseases [
69
]. Additionally, caries-associated bacteria did not strongly
contribute to the overall classification within this study, as shown by the RF modeling.
For the qPCR detection, there are two potential causes for the missing values, the absence
or small quantity of analyte in the patient saliva or inadequate assay performance. As
previously described, all the primer and probe sets were validated using genomic reference
DNA using standards ranging in amount from 0.1 pg to 10 ng DNA. The qPCR was repeated
on three consecutive days, preparing all the reagents from scratch each time, with the MIQE
guidelines being followed throughout the study [
70
]. The low abundance of Aggregatibacter
actinomycetemcomitans in the general population has already been reported [
71
]. However,
it appears that the assay performance for Streptococcus sobrinus and Lactobacilli may have
been impaired by the presence of salivary inhibitors. Hence, it cannot be fully excluded
that those bacteria are important drivers in the development or identification of dental
caries. The total human protease, matrix metalloproteinase, and metallopeptidase inhibitor
levels did not enhance the predictability (caries, gingivitis, or healthy) within this study.
However, their major contribution to the prediction of caries was not to be expected, as
the proteases MMP-8 and MMP-9 are part of a highly complex “protease web”, which is
mainly associated with destructive periodontal disease [72].
The lack of research in the field of salivary caries signatures to date might be based on
the nature of current caries diagnostics, which are primarily applied with visual examina-
tions including optical caries detection devices, tactile assessments, and radiographs [
73
,
74
].
The molecular analyses presented in this publication investigated diverse salivary protein
biomarkers and bacterial population levels, and the results could potentially be used to
increase the sensitivity of caries detection, as well as to improve caries prevention strategies.
A full description of the underlying pathways and involved mechanisms was not possible
since this study was first concerned with the screening of potential biomarkers that are
able to predict dental caries. The data suggest a potential association between JAK/STAT
signaling and linked Th1/Th2 immune responses; however, further research should be
conducted in this direction.
Future experiments should also test the universal applicability of the four biomarkers
to other patient groups, e.g., patients with periodontitis, heavy smokers, or pregnant
and lactating women. Furthermore, potential effects on the oral immune response and
microbiome following the consumption of vaporizers and e-cigarettes should be addressed.
J. Pers. Med. 2021,11, 235 15 of 18
5. Conclusions
The current research identified four biomarkers (IL-4, IL-13, IL-2-RA, and eotaxin/CCL11)
that enabled the correct diagnosis of dental caries in 37 out of 38 patients using RF analy-
sis. Ten subjects were followed over a period of six months, with their oral health status
being clinically assessed and compared to biomarker-based RF predictions. We suggest a
further validation of the four biomarkers (IL-4, IL-13, IL-2-RA, and eotaxin/CCL11) in the
context of JAK/STAT signaling and dental caries. Our results highlight the importance of
additional sensitive molecular assays, acting in a way complementary to existing clinical
assessment methodologies and enabling a holistic and personalized approach to caries
detection and therapy. Biomarker assays may facilitate this approach and allow dentists to
accurately keep track of patients’ recovery towards healthy oral microenvironments. This
research has laid the foundation for the development of simple and economically feasible
saliva-based diagnostic tests aimed at assessing the presence/absence of dental caries in a
personalized manner.
Supplementary Materials:
The following are available online at https://www.mdpi.com/2075-442
6/11/3/235/s1. Table S1: Overview of Cytokine 30-Plex Human Panel, Table S2: Upper and lower
detection limits of quantification by assay, Table S3: Compiled dataset including clinical assessment
data of all subjects.
Author Contributions:
Conceptualization, P.N.P., C.H., D.B.W., K.M., J.R.P. and N.B.; Data curation,
P.N.P., C.H., D.B.W. and J.S.J.; Formal analysis, P.N.P., C.H., D.B.W. and J.S.J.; Funding acquisition,
K.M., T.A., G.N.B., J.R.P. and N.B.; Investigation, P.N.P. and C.H.; Methodology, P.N.P., C.H., D.B.W.
and N.B.; Project administration, P.N.P., C.H., K.M., T.A., G.N.B., J.R.P., F.J.W. and N.B.; Resources,
K.M., T.A., K.B., J.S.J., W.E.K., M.K., J.R.P., P.R.S., T.T. and N.B.; Software, D.B.W.; Supervision,
K.M., T.A. and J.R.P.; Validation, P.N.P., C.H. and D.B.W.; Visualization, P.N.P., C.H. and D.B.W.;
Writing—original draft, P.N.P. and C.H.; Writing—review & editing, P.N.P., C.H., D.B.W., K.M., T.A.,
G.N.B., J.P.H., J.S.J., W.E.K., M.K., P.K., J.R.P., P.R.S., T.T., F.J.W. and N.B. All authors have read and
agreed to the published version of the manuscript.
Funding:
This research was funded by the European Union’s Horizon 2020 research and innovation
program, grant number 633780 (“DIAGORAS” project).
Institutional Review Board Statement:
The study was conducted according to the guidelines of the
Declaration of Helsinki, and approved by the local Swiss ethics committee (BASEC-no. 2016-00435,
date of approval: 09.01.2016).
Informed Consent Statement:
Informed consent was obtained from all the subjects involved in
the study.
Data Availability Statement:
The data are contained within the article or Supplementary Materials.
The data presented in this study are available in Supplementary Data S3.
Acknowledgments:
We would like to thank Helga Lüthi-Schaller for her excellent laboratory and
technical support during the project and experiments and are grateful for her constructive ideas.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Chapple, I.L.C.; Van Der Weijden, F.; Doerfer, C.; Herrera, D.; Shapira, L.; Polak, D.; Madianos, P.; Louropoulou, A.; Machtei, E.;
Donos, N.; et al. Primary prevention of periodontitis: Managing gingivitis. J. Clin. Periodontol. 2015,42, S71–S76. [CrossRef]
2.
Sanz, M.; Beighton, D.A.; Curtis, M.; Cury, J.A.; Dige, I.; Dommisch, H.; Ellwood, R.; Giacaman, R.A.; Herrera, D.; Herzberg,
M.C.; et al. Role of microbial biofilms in the maintenance of oral health and in the development of dental caries and periodontal
diseases. Consensus report of group 1 of the Joint EFP/ORCA workshop on the boundaries between caries and periodontal
disease. J. Clin. Periodontol 2017,44, S5–S11. [CrossRef]
3. Socransky, S.S.; Haffajee, A.D. Periodontal microbial ecology. Periodontol. 2000 2005,38, 135–187. [CrossRef] [PubMed]
4.
Marsh, P.D. The significance of maintaining the stability of the natural microflora of the mouth. Br. Dent. J.
1991
,
171, 174–177. [CrossRef]
5.
Dewhirst, F.E.; Chen, T.; Izard, J.; Paster, B.J.; Tanner, A.C.R.; Yu, W.-H.; Lakshmanan, A.; Wade, W.G. The Human Oral
Microbiome. J. Bacteriol. 2010,192, 5002–5017. [CrossRef]
J. Pers. Med. 2021,11, 235 16 of 18
6.
Marsh, P. Microbial Ecology of Dental Plaque and its Significance in Health and Disease. Adv. Dent. Res.
1994
,8, 263–271.
[CrossRef] [PubMed]
7.
Shungin, D.; Haworth, S.; Divaris, K.; Agler, C.S.; Kamatani, Y.; Lee, M.K.; Grinde, K.; Hindy, G.; Alaraudanjoki, V.;
Pesonen, P.; et al.
Genome-wide analysis of dental caries and periodontitis combining clinical and self-reported data. Nat.
Commun. 2019,10, 1–13. [CrossRef]
8.
Zaura, E.; Cate, J.T. Dental Plaque as a Biofilm: A Pilot Study of the Effects of Nutrients on Plaque pH and Dentin Demineralization.
Caries Res. 2003,38, 9–15. [CrossRef]
9.
Rosier, B.; Van Loveren, C.; Zaura, E.; Loos, B.; Keijser, B.; Crielaard, W.; Lagerweij, M. Caries Incidence in a Healthy Young Adult
Population in Relation to Diet. JDR Clin. Transl. Res. 2016,2, 142–150. [CrossRef]
10.
Hujoel, P.P.; Lingström, P. Nutrition, dental caries and periodontal disease: A narrative review. J. Clin. Periodontol.
2017
,
44, S79–S84. [CrossRef] [PubMed]
11. Selwitz, R.H.I.; Ismail, A.; Pitts, N.B. Dental caries. Lancet 2007,369, 51–59. [CrossRef]
12.
Paqué, P.N.; Herz, C.; Jenzer, J.S.; Wiedemeier, D.B.; Attin, T.; Bostanci, N.; Belibasakis, G.N.; Bao, K.; Körner, P.;
Fritz, T.; et al.
Microbial Analysis of Saliva to Identify Oral Diseases Using a Point-of-Care Compatible qPCR Assay.
J. Clin. Med. 2020,9, 2945. [CrossRef]
13.
Trombelli, L.; Tatakis, D.N.; Scapoli, C.; Bottega, S.; Orlandini, E.; Tosi, M. Modulation of clinical expression of plaque-
induced gingivitis. II. Identification of “high-responder” and “low-responder” subjects. J. Clin. Periodontol.
2004
,31, 239–252.
[CrossRef] [PubMed]
14.
Trombelli, L.; Scapoli, C.; Orlandini, E.; Tosi, M.; Bottega, S.; Tatakis, D.N. Modulation of clinical expression of plaque-induced
gingivitis. III. Response of “high responders” and “low responders” to therapy. J. Clin. Periodontol.
2004
,31, 253–259.
[CrossRef] [PubMed]
15. Larsen, T.; Fiehn, N.-E. Dental biofilm infections—An update. APMIS 2017,125, 376–384. [CrossRef] [PubMed]
16.
Hajishengallis, G.; Darveau, R.P.; Curtis, M.A. The keystone-pathogen hypothesis. Nat. Rev. Microbiol.
2012
,10, 717–725.
[CrossRef] [PubMed]
17. Löe, H.; Theilade, E.; Jensen, S.B. Experimental Gingivitis in Man. J. Periodontol. 1965,36, 177–187. [CrossRef]
18.
Bostanci, N.; Silbereisen, A.; Bao, K.; Grossmann, J.; Nanni, P.; Fernandez, C.; Nascimento, G.G.; Belibasakis, G.N.; Lopez, R.
Salivary proteotypes of gingivitis tolerance and resilience. J. Clin. Periodontol. 2020,47, 1304–1316. [CrossRef] [PubMed]
19.
Offenbacher, S.; Barros, S.; Mendoza, L.; Mauriello, S.; Preisser, J.; Moss, K.; De Jager, M.; Aspiras, M. Changes in gingival
crevicular fluid inflammatory mediator levels during the induction and resolution of experimental gingivitis in humans. J. Clin.
Periodontol. 2010,37, 324–333. [CrossRef]
20.
Bostanci, N.; Ramberg, P.; Wahlander, Å.; Grossman, J.; Jönsson, D.; Barnes, V.M.; Papapanou, P.N. Label-Free Quantitative
Proteomics Reveals Differentially Regulated Proteins in Experimental Gingivitis. J. Proteome Res. 2013,12, 657–678. [CrossRef]
21.
Gupta, G. Gingival crevicular fluid as a periodontal diagnostic indicator- I: Host derived enzymes and tissue breakdown products.
J. Med. Life 2012,5, 390–397.
22.
Farges, J.-C.; Keller, J.-F.; Carrouel, F.; Durand, S.H.; Romeas, A.; Bleicher, F.; Lebecque, S.; Staquet, M.-J. Odontoblasts in the
dental pulp immune response. J. Exp. Zool. Part B Mol. Dev. Evol. 2009,312B, 425–436. [CrossRef] [PubMed]
23.
Giannobile, W.V.; Beikler, T.; Kinney, J.S.; Ramseier, C.A.; Morelli, T.; Wong, D.T. Saliva as a diagnostic tool for periodontal disease:
Current state and future directions. Periodontol. 2000 2009,50, 52–64. [CrossRef] [PubMed]
24.
Kc, S.; Wang, X.Z.; Gallagher, J.E.; Sukriti, K. Diagnostic sensitivity and specificity of host-derived salivary biomarkers in
periodontal disease amongst adults: Systematic review. J. Clin. Periodontol. 2019,47, 289–308. [CrossRef] [PubMed]
25.
Greenwood, D.; Afacan, B.; Emingil, G.; Bostanci, N.; Belibasakis, G.N. Salivary Microbiome Shifts in Response to Periodontal
Treatment Outcome. Proteom. Clin. Appl. 2020,14, e2000011. [CrossRef]
26.
Lundmark, A.; Hu, Y.O.O.; Huss, M.; Johannsen, G.; Andersson, A.F.; Yucel-Lindberg, T. Identification of Salivary Microbiota and Its
Association With Host Inflammatory Mediators in Periodontitis. Front. Cell. Infect. Microbiol. 2019,9, 216. [CrossRef] [PubMed]
27.
Korte, D.L.; Kinney, J. Personalized medicine: An update of salivary biomarkers for periodontal diseases. Periodontol. 2000
2015
,
70, 26–37. [CrossRef]
28.
Jaedicke, K.M.; Preshaw, P.M.; Taylor, J.J. Salivary cytokines as biomarkers of periodontal diseases. Periodontol. 2000
2015
,
70, 164–183. [CrossRef]
29.
Kinney, J.; Morelli, T.; Braun, T.; Ramseier, C.; Herr, A.; Sugai, J.; Shelburne, C.; Rayburn, L.; Singh, A.; Giannobile, W.
Saliva/Pathogen Biomarker Signatures and Periodontal Disease Progression. J. Dent. Res. 2011,90, 752–758. [CrossRef]
30.
Silbereisen, A.; Alassiri, S.; Bao, K.; Grossmann, J.; Nanni, P.; Fernandez, C.; Tervahartiala, T.; Nascimento, G.G.; Belibasakis,
G.N.; Heikkinen, A.; et al. Label-Free Quantitative Proteomics versus Antibody-Based Assays to Measure Neutrophil-Derived
Enzymes in Saliva. Proteom. Clin. Appl. 2019,14. [CrossRef]
31.
Akcalı, A.; Bostanci, N.; Özçaka, Ö.; Öztürk-Ceyhan, B.; Gümü¸s, P.; Buduneli, N.; Belibasakis, G.N. Association between Polycystic
Ovary Syndrome, Oral Microbiota and Systemic Antibody Responses. PLoS ONE 2014,9, e108074. [CrossRef]
32.
Ebersole, J.L.; Nagarajan, R.; Akers, D.; Miller, C.S. Targeted salivary biomarkers for discrimination of periodontal health and
disease(s). Front. Cell. Infect. Microbiol. 2015,5, 62. [CrossRef]
33.
Sexton, W.M.; Lin, Y.; Kryscio, R.J.; Dawson, D.R.; Ebersole, J.L.; Miller, C.S. Salivary biomarkers of periodontal disease in
response to treatment. J. Clin. Periodontol. 2011,38, 434–441. [CrossRef]
J. Pers. Med. 2021,11, 235 17 of 18
34.
Rechenberg, D.-K.; Galicia, J.C.; Peters, O.A. Biological Markers for Pulpal Inflammation: A Systematic Review. PLoS ONE
2016
,
11, e0167289. [CrossRef] [PubMed]
35.
Martins, C.; Buczynski, A.K.; Maia, L.C.; Siqueira, W.L.; Castro, G.F.B.D.A. Salivary proteins as a biomarker for dental caries—A
systematic review. J. Dent. 2013,41, 2–8. [CrossRef] [PubMed]
36.
Farges, J.-C.; Alliot-Licht, B.; Renard, E.; Ducret, M.; Gaudin, A.; Smith, A.J.; Cooper, P.R. Dental Pulp Defence and Repair
Mechanisms in Dental Caries. Mediat. Inflamm. 2015,2015, 1–16. [CrossRef] [PubMed]
37.
Ayad, M.; Van Wuyckhuyse, B.; Minaguchi, K.; Raubertas, R.; Bedi, G.; Billings, R.; Bowen, W.; Tabak, L. The association of basic
proline-rich peptides from human parotid gland secretions with caries experience. J. Dent. Res. 2000,79, 976–982. [CrossRef]
38.
Tulunoglu, O.; Demirtas, S. Total antioxidant levels of saliva in children related to caries, age, and gender. Int. J. Paediatr. Dent.
2006,16, 186–191. [CrossRef]
39.
Roa, N.S.; Chaves, M.; Gómez, M.; Jaramillo, L.M. Association of salivary proteins with dental caries in a Colombian population.
Acta Odontol Latinoam 2008,21, 69–75.
40.
Yoshizawa, J.M.; Schafer, C.A.; Schafer, J.J.; Farrell, J.J.; Paster, B.J.; Wong, D.T.W. Salivary Biomarkers: Toward Future Clinical
and Diagnostic Utilities. Clin. Microbiol. Rev. 2013,26, 781–791. [CrossRef] [PubMed]
41. Streckfus, C.F.; Bigler, L.R. Saliva as a diagnostic fluid. Oral Dis. 2002,8, 69–76. [CrossRef]
42.
Mitsakakis, K.; Stumpf, F.; Strohmeier, O.; Klein, V.; Mark, D.; Von Stetten, F.; Peham, J.R.; Herz, C.; Paqué, P.;
Wegehaupt, F.; et al.
Chair/bedside diagnosis of oral and respiratory tract infections, and identification of antibiotic resistances for personalised
monitoring and treatment. Stud. Health Technol. Inform. 2016,224, 61–66.
43.
Belibasakis, G.N.; Bostanci, N.; Marsh, P.D.; Zaura, E. Applications of the oral microbiome in personalized dentistry. Arch. Oral
Biol. 2019,104, 7–12. [CrossRef]
44. Proctor, G.B. The physiology of salivary secretion. Periodontol. 2000 2016,70, 11–25. [CrossRef]
45.
Dige, I.; Schlafer, S.; Nyvad, B. Difference in initial dental biofilm accumulation between night and day. Acta Odontol. Scand.
2011
,
70, 441–447. [CrossRef]
46. Navazesh, M.; Kumar, S.K. Measuring salivary flow. J. Am. Dent. Assoc. 2008,139, 35S–40S. [CrossRef]
47.
4Hong, I.; Pae, H.-C.; Song, Y.W.; Cha, J.-K.; Lee, J.-S.; Paik, J.-W.; Choi, S.-H. Oral Fluid Biomarkers for Diagnosing Gingivitis in
Human: A Cross-Sectional Study. J. Clin. Med. 2020,9, 1720. [CrossRef]
48.
Costantini, E.; Sinjari, B.; Piscopo, F.; Porreca, A.; Reale, M.; Caputi, S.; Murmura, G. Evaluation of Salivary Cytokines and
Vitamin D Levels in Periodontopathic Patients. Int. J. Mol. Sci. 2020,21, 2669. [CrossRef] [PubMed]
49.
Verhulst, M.J.L.; Teeuw, W.J.; Bizzarro, S.; Muris, J.; Su, N.; Nicu, E.A.; Nazmi, K.; Bikker, F.J.; Loos, B.G. A rapid, non-invasive
tool for periodontitis screening in a medical care setting. BMC Oral Health 2019,19, 87. [CrossRef]
50.
Nucleotide: Bethesda (MD): National Library of Medicine (US), National Center for Biotechnology Information. 1988. Available
online: https://www.ncbi.nlm.nih.gov/nucleotide/ (accessed on 6 May 2017).
51.
Untergasser, A.; Cutcutache, I.; Koressaar, T.; Ye, J.; Faircloth, B.C.; Remm, M.; Rozen, S.G. Primer3—new capabilities and
interfaces. Nucleic Acids Res. 2012,40, e115. [CrossRef] [PubMed]
52.
Assembly: Bethesda (MD): National Library of Medicine (US), National Center for Biotechnology Information. 2012. Available
online: https://www.ncbi.nlm.nih.gov/nucleotide/ (accessed on 6 May 2017).
53.
Wei, R.; Wang, J.; Su, M.; Jia, E.; Chen, S.; Chen, T.; Ni, Y. Missing Value Imputation Approach for Mass Spectrometry-based
Metabolomics Data. Sci. Rep. 2018,8, 1–10. [CrossRef]
54.
Paluszynska, A.; Biecek, P.; Jiang, Y. randomForestExplainer: Explaining and Visualizing Random Forests in Terms of Variable
Importance. R package version 0.10.1. Available online: https://CRAN.R-project.org/package=randomForestExplainer (accessed
on 9 March 2020).
55.
Ehrlinger, J. ggRandomForests: Visually Exploring Random Forests. R package version 2.0.1. Available online: https://CRAN.R-
project.org/package=ggRandomForests (accessed on 17 March 2021).
56.
R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing: Vienna,
Austria, 2018; Available online: https://www.R-project.org (accessed on 4 February 2021).
57. Wickham, H. ggplot2: Elegant Graphics for Data Analysis; Springer: New York, NY, USA, 2016; p. 213.
58.
Pohlert, T. PMCMRplus: Calculate Pairwise Multiple Comparisons of Mean Rank Sums Extended. R package version 1.4.4. 2020.
Available online: https://CRAN.R-project.org/package=PMCMRplus (accessed on 4 February 2021).
59. Liaw, A.; Wiener, M. Classification and Regression by randomForest. R News 2002,2, 18–22.
60. Lê, S.; Josse, J.; Husson, F. FactoMineR: AnRPackage for Multivariate Analysis. J. Stat. Softw. 2008,25, 1–18. [CrossRef]
61. Paul, W.E. History of interleukin-4. Cytokine 2015,75, 3–7. [CrossRef] [PubMed]
62.
Minty, A.; Chalon, P.; Derocq, J.-M.; Dumont, X.; Guillemot, J.-C.; Kaghad, M.; Labit, C.; Leplatois, P.; Liauzun, P.;
Miloux, B.; et al.
lnterleukin-13 is a new human lymphokine regulating inflammatory and immune responses. Nat. Cell Biol.
1993,362, 248–250. [CrossRef]
63.
Zurawski, G.; De Vries, J.E. Interleukin 13, an interleukin 4-like cytokine that acts on monocytes and B cells, but not on T cells.
Immunol. Today 1994,15, 19–26. [CrossRef]
64.
Bhattacharjee, A.; Shukla, M.; Yakubenko, V.P.; Mulya, A.; Kundu, S.; Cathcart, M.K. IL-4 and IL-13 employ discrete
signaling pathways for target gene expression in alternatively activated monocytes/macrophages. Free. Radic. Biol. Med.
2013,54, 1–16. [CrossRef]
J. Pers. Med. 2021,11, 235 18 of 18
65.
Malabarba, M.G.; Rui, H.; Deutsch, H.H.J.; Chung, J.; Kalthoff, F.S.; Farrar, W.L.; Kirken, R.A. Interleukin-13 is a potent
activator of JAK3 and STAT6 in cells expressing interleukin-2 receptor-
γ
and interleukin-4 receptor-
α
.Biochem. J.
1996
,
319, 865–872. [CrossRef]
66.
Hosokawa, Y.; Hosokawa, I.; Shindo, S.; Ozaki, K.; Matsuo, T. IL-4 Modulates CCL11 and CCL20 Productions from IL-1
β
-
Stimulated Human Periodontal Ligament Cells. Cell. Physiol. Biochem. 2016,38, 153–159. [CrossRef] [PubMed]
67.
Weijden, G.A.; Timmerman, M.F.; Nijboer, A.; Reijerse, E.; Velden, U. Comparison of different approaches to assess bleeding on
probing as indicators of gingivitis. J. Clin. Periodontol. 1994,21, 589–594. [CrossRef] [PubMed]
68.
Lopez, R.; Hujoel, P.; Belibasakis, G.N. On putative periodontal pathogens: An epidemiological perspective. Virulence
2015
,
6, 249–257. [CrossRef] [PubMed]
69.
Bostanci, N.; Bao, K.; Greenwood, D.; Silbereisen, A.; Belibasakis, G.N. Periodontal disease: From the lenses of light microscopy
to the specs of proteomics and next-generation sequencing. Adv. Clin. Chem. 2019,93, 263–290. [CrossRef] [PubMed]
70.
Bustin, S.A.; Benes, V.; Garson, J.A.; Hellemans, J.; Huggett, J.; Kubista, M.; Mueller, R.; Nolan, T.; Pfaffl, M.W.; Shipley, G.L.; et al.
The MIQE Guidelines: Minimum Information for Publication of Quantitative Real-Time PCR Experiments. Clin. Chem.
2009
,
55, 611–622. [CrossRef] [PubMed]
71.
Fine, D.H.; Markowitz, K.; Furgang, D.; Velliyagounder, K. Aggregatibacter actinomycetemcomitansas an Early Colonizer of Oral
Tissues: Epithelium as a Reservoir? J. Clin. Microbiol. 2010,48, 4464–4473. [CrossRef]
72.
Cavalla, F.; Hernandez-Rios, P.; Sorsa, T.; Biguetti, C.; Hernandez, M. Matrix Metalloproteinases as Regulators of Periodontal
Inflammation. Int. J. Mol. Sci. 2017,18, 440. [CrossRef]
73.
Schwendicke, F.; Brouwer, F.; Paris, S.; Stolpe, M. Detecting Proximal Secondary Caries Lesions. J. Dent. Res.
2015
,95, 152–159.
[CrossRef] [PubMed]
74. Brouwer, F.; Askar, H.; Paris, S.; Schwendicke, F. Detecting Secondary Caries Lesions. J. Dent. Res. 2016,95, 143–151. [CrossRef]
... It is the dynamic changes over time of bacteria among the oral microbiota, rather than their mere presence/absence, that drives the dysbiotic changes that lead to oral disease [72]. The fact that the OralDisk has demonstrated its capacity to analyze oral bacteria provides a basis upon which the platform can be implemented in time-course studies associated with patient monitoring, similar to the rationale reported by Paqué et al. [73]. In such future studies, we will have the opportunity to proceed to the quantification of bacteria, using OralDisk-derived calibration curves to convert Cq values into numeric bacterial loads. ...
... This will increase the interoperability between immunological and microbiological diagnostic outputs using this platform [76]. Subsequently, by using combined computational technologies and the OralDisk device, new diagnostic predictive models of disease development and progression may be generated [73]. Such developments will enable evidence-based pre-and post-treatment monitoring of a range of oral (OralDisk) and non-oral (application-specific LabDisk) diseases. ...
Article
Full-text available
Periodontitis and dental caries are two major bacterially induced, non-communicable diseases that cause the deterioration of oral health, with implications in patients' general health. Early, precise diagnosis and personalized monitoring are essential for the efficient prevention and management of these diseases. Here, we present a disk-shaped microfluidic platform (OralDisk) compatible with chair-side use that enables analysis of non-invasively collected whole saliva samples and molecular-based detection of ten bacteria: seven periodontitis-associated (Aggregatibacter actinomycetemcomitans, Campylobacter rectus, Fusobacterium nucleatum, Prevotella intermedia, Porphyromonas gingivalis, Tannerella forsythia, Treponema denticola) and three caries-associated (oral Lactobacilli, Streptococcus mutans, Streptococcus sobrinus). Each OralDisk test required 400 µL of homogenized whole saliva. The automated workflow included bacterial DNA extraction, purification and hydrolysis probe real-time PCR detection of the target pathogens. All reagents were pre-stored within the disk and sample-to-answer processing took < 3 h using a compact, customized processing device. A technical feasibility study (25 OralDisks) was conducted using samples from healthy, periodontitis and caries patients. The comparison of the OralDisk with a lab-based reference method revealed a ~90% agreement amongst targets detected as positive and negative. This shows the OralDisk's potential and suitability for inclusion in larger prospective implementation studies in dental care settings.
... One of the diseases caused by S. mutans is dental caries. There are several factors that cause dental caries to get worse including sugar, saliva, and also putrefactive bacteria [25][26][27]. In addition, the growth of bacteria in the mouth and forming biofilms is caused by several factors, namely saliva which plays a role in modulating the plaque layer on the teeth, the temperature in the environment around the mouth in the range of 35-36°C, and pH 6.75-7.25 [28,29]. ...
Article
Full-text available
Based on data from The Global Burden of Disease Study in 2016, dental and oral health problems, especially dental caries, are a disease experienced by almost half of the world’s population (3.58 billion people). One of the main causes of dental caries is the pathogenesis of Streptococcus mutans. Prevention can be achieved by controlling S. mutans using an antibacterial agent. The most commonly used antibacterial for the treatment of dental caries is chlorhexidine. However, long-term use of chlorhexidine has been reported to cause resistance and some side effects. Therefore, the discovery of a natural antibacterial agent is an urgent need. A natural antibacterial agent that can be used are herbal medicines derived from medicinal plants. Piper crocatum Ruiz and Pav has the potential to be used as a natural antibacterial agent for treating dental and oral health problems. Several studies reported that the leaves of P. crocatum Ruiz and Pav contain secondary metabolites such as essential oils, flavonoids, alkaloids, terpenoids, tannins, and phenolic compounds that are active against S. mutans. This review summarizes some information about P. crocatum Ruiz and Pav, various isolation methods, bioactivity, S. mutans bacteria that cause dental caries, biofilm formation mechanism, antibacterial properties, and the antibacterial mechanism of secondary metabolites in P. crocatum Ruiz and Pav.
... This article is protected by copyright. All rights reserved Paqué et al. 2021), or qualitative detection of antibiotic resistance genes (Belibasakis et al. 2020). Saliva sampling may also conveniently replace the uncomfortable nasal or oropharyngeal sampling for select applications, such as SARS-CoV-2 RNA detection, with high specificity and sensitivity (Atieh et al. 2021). ...
Article
Full-text available
The authors of this narrative review aimed to address various experimental methods and make recommendations for how research should move forward in the context of studying biomarkers in clinical Endodontic research. The approach adopted is exemplified using two prominent clinical problems, namely a) the "reversible" versus "irreversible" pulpitis conundrum and b) persistent idiopathic dentoalveolar pain (PIDAP). Pulpitis under deep caries or dentinal cracks is understood from a histological perspective, but clinical assessment tools to indicate irreversibly inflamed aspects of the dental pulp are elusive. PIDAP, on the other hand, is a diagnosis of exclusion; its pathophysiology is complex and not understood sufficiently to avoid unnecessary dental treatments. This review addresses how diagnostic biomarkers could further our understanding of those and other clinical problems, and how issues can be tackled from a methodological point of view. Hence, different methodological approaches to identify suitable diagnostic biomarker(s) or use known biomarkers are presented. The importance of asking a relevant research question, collecting the most suitable fluid, and using the ideal collection vehicle for the research question under investigation is discussed based on the defined clinical problems.
... Paqué et al. [15] investigated the potential of salivary bacterial and protein markers for evaluating the disease status in healthy individuals with caries. Saliva samples from caries and gingivitis-free individuals (n = 18), patients with deep caries lesions (n = 38) was collected and analysed for 44 candidate biomarkers namely: selected oral bacteria, growth factors, chemokines and proteolytic enzymes among others. ...
Article
Full-text available
Dental Caries are one of the most prevalent chronic diseases around the globe. Detecting carious lesions is a challenging task. Conventional computer aided diagnosis and detection methods in the past have heavily relied on the visual inspection of teeth. These methods are only effective on large and clearly visible caries on affected teeth. Conventional methods have been limited in performance due to the complex visual characteristics of dental caries images, which consist of hidden or inaccessible lesions. The early detection of dental caries is an important determinant for treatment and benefits much from the introduction of new tools, such as dental radiography. In this paper, we propose a deep learning-based technique for dental caries detection namely: blob detection. The proposed technique automatically detects hidden and inaccessible dental caries lesions in bitewing radio-graphs. The approach employs data augmentation to increase the number of images in the data set to have a total of 11,114 dental images. Image pre-processing on the data set was through the use of Gaussian blur filters. Image segmentation was handled through thresholding, erosion and dilation morphology, while image boundary detection was achieved through active contours method. Furthermore, the deep learning based network through the sequential model in Keras extracts features from the images through blob detection. Finally, a convexity threshold value of 0.9 is introduced to aid in the classification of caries as either present or not present. The process of detection and classifying dental caries achieved the results of 97% and 96% for the precision and recall values, respectively.
... The prevalence and damage generated by these bacteria are influenced by host protective factors, such as immune competence, the production and buffering capacity of saliva, and hygiene and dietary habits [5]. Mechanisms have been sought to diagnose the risk of caries based on these factors using software such as Cariogram [6] or through the identification of biomarkers [7,8,9]. However, a significant amount of new studies are required to evaluate the validity of caries biomarkers, among other things because tooth decay is a multi-factorial disease with genetic, environmental and microbial effects playing a role, and because of a large inter-individual variability that renders the task of finding universal caries markers [10]. ...
Article
Full-text available
Background: Electrolytes, proteins, and other salivary molecules play an important role in tooth integrity and can serve as biomarkers associated with caries. Objective: To determine the concentration of potential biomarkers in children without caries (CF) and children with caries (CA). Methods: Unstimulated saliva was collected, and the biomarkers quantified in duplicate, using commercial Enzyme Linked Immunosorbent Assay (ELISA) kits to determine IgA, fibronectin, cathelicidin LL-37, and statherin levels, as well as colorimetric tests to detect formate and phosphate. Results: Significantly higher concentrations of statherin was detected in the CF group (Median: 94,734.6; IQR: 92,934.6–95,113.7) compared to the CA2 group (90,875.0; IQR: 83,580.2–94,633.4) (p = 0.03). Slightly higher median IgA (48,250.0; IQR: 31,461.9–67,418.8) and LL-37 levels (56.1; IQR 43.6–116.2) and a lower concentration of formate were detected in the CF group (0.02; IQR 0.0034–0.15) compared to the group with caries (IgA: 37,776.42; IQR: 33,383.9–44,128.5; LL-37: 46.3; IQR: 40.1011–67.7; formate: 0.10; IQR: 0.01–0.18), but these differences were not statistically significant. Conclusion: The fact that these compounds have been identified as good markers for caries among European adults highlights the difficulty of identifying universal biomarkers that are applicable to all ages or to different populations.
Article
Full-text available
Oral health is maintained by a healthy microbiome, which can be monitored by state-of-the art diagnostics. Therefore, this study evaluated the presence and quantity of ten oral disease-associated taxa (P. gingivalis, T. forsythia, T. denticola, F. nucleatum, C. rectus, P. intermedia, A. actinomycetemcomitans, S. mutans, S. sobrinus, oral associated Lactobacilli) in saliva and their clinical status association in 214 individuals. Upon clinical examination, study subjects were grouped into healthy, caries and periodontitis and their saliva was collected. A highly specific point-of-care compatible dual color qPCR assay was developed and used to study the above-mentioned bacteria of interest in the collected saliva. Assay performance was compared to a commercially available microbial reference test. Eight out of ten taxa that were investigated during this study were strong discriminators between the periodontitis and healthy groups: C. rectus, T. forsythia, P. gingivalis, S. mutans, F. nucleatum, T. denticola, P. intermedia and oral Lactobacilli (p < 0.05). Significant differentiation between the periodontitis and caries group microbiome was only shown for S. mutans (p < 0.05). A clear distinction between oral health and disease was enabled by the analysis of quantitative qPCR data of target taxa levels in saliva.
Article
Full-text available
Aim This study aimed to characterize the salivary proteome during the induction and resolution of gingival inflammation in the course of human experimental gingivitis (EG), and to cluster the proteomic profiles based on the clinically‐defined ‘slow’ and ‘fast’ response patterns. Materials and methods A total of 50 unstimulated whole saliva were obtained from the EG model which was induced over 21 days (day 0, 7, 14, 21), followed by a two‐week resolution phase (day 35). Label‐free quantitative proteomics using liquid chromatography‐tandem mass spectrometry was applied. Regulated proteins were subject to gene ontology enrichment analysis. Results A total of 804 human proteins were quantified by ≥2 peptides. Principal component analysis depicted significant differences between ‘fast’ and ‘slow’ responders. Despite gingival and plaque scores being similar at baseline among the two groups, ‘fast’ responders presented with 48 proteins that were at >4‐fold higher levels than ‘slow’ responders. These up‐regulated proteins showed enrichment in “antigen presentation” and “proteolysis”. Conclusions Together, these findings highlight the utility of integrative systems‐level quantitative proteomic approaches to unravel the molecular basis of ‘salivary proteotypes’ associated with gingivitis dubbed as ‘fast’ and ‘slow’ responders. Hence, these differential responses may help prognosticate individual susceptibility to gingival inflammation.
Article
Full-text available
Diagnoses based on oral fluid biomarkers have been introduced to overcome limitations of periodontal probe-based diagnoses. Diagnostic ability of certain biomarkers for periodontitis have been identified and widely studied, however, such studies targeting gingivitis is scarce. The aims of this study were to determine and compare the efficacies and accuracies of eight biomarkers in diagnosing gingivitis with the aid of receiver operating characteristic (ROC) curves. The probing depth (PD), clinical attachment loss (CAL), bleeding on probing (BOP), gingival index (GI), and plaque index (PI) were examined in 100 participants. Gingival crevicular fluid was collected using paper points, and whole-saliva samples were collected using cotton roll. Samples were analyzed using enzyme-linked immunosorbent assay kits for the different biomarkers. The levels of matrix metalloproteinase (MMP)-8, MMP-9, lactoferrin, cystatin C, myeloperoxidase (MPO), platelet-activating factor, cathepsin B, and pyridinoline cross-linked carboxyterminal telopeptide of type I collagen were analyzed. MPO and MMP-8 levels in saliva were strongly correlated with gingivitis, with Pearson’s correlation coefficients of 0.399 and 0.217, respectively. The area under the curve (AUC) was largest for MMP-8, at 0.814, followed by values of 0.793 and 0.777 for MPO and MMP-9, respectively. The clinical parameters of GI and PI showed strong correlations and large AUC values, whereas PD and CAL did not. MMP-8 and MPO were found to be effective for diagnosing gingivitis. Further investigations based on the results of this study may identify clinically useful biomarkers for the accurate and early detection of gingivitis.
Article
Full-text available
Periodontal disease (PD) is an inflammatory condition of the tissues supporting the teeth, which is widespread among the adult population. Evidence shows a relationship between PD and vitamin D levels, which is involved in the regulation of bone metabolism, mineral homeostasis, and inflammatory response. This study aimed to perform a simultaneous evaluation of inflammatory mediators and vitamin D levels in saliva in periodontopathic patients to better understand their role in periodontal disease. In this observational study, clinical periodontal parameter examination was performed for each patient. Moreover, the saliva levels of 25(OH)D3, TGFβ, IL-35, IL-17A, and MMP9 were evaluated using an ELISA assay. An increase in TGFβ, IL-35, MMP9, and IL-17A salivary levels and a reduction in 25(OH)D3 levels were observed in periodontopathic patients with respect to the healthy controls. The present study revealed significant positive correlation between cytokines and highly negative correlation between 25(OH)D3 and salivary cytokine levels. Further studies are needed to better understand if salivary cytokines and vitamin D evaluation may represent a new approach for detection and prevention of progressive diseases, such as PD.
Article
Full-text available
Purpose: Periodontitis is linked to a localized dysbiotic microbial shift. This trending may often not be evident due to deep taxonomic changes of low abundance organisms and lack of consideration of variations in the treatment response. By using next generation sequencing this study aimed to evaluate the salivary microbiome dynamics of periodontal treatment and the implication of its outcome. Experimental design: Patients with generalized aggressive periodontitis were treated non‐surgically and followed up for 6 months. Saliva was collected for microbiome profiling by next generation sequencing and diversity analysis, as well as qPCR. The treatment outcome on the first follow‐up was also considered. Results: Clinical parameters were significantly improved following treatment, but with no accompanying relative abundance changes on the phylum, genus and species levels, or diversity indices. Distinctive differences were observed on species level when the sensitive qPCR was used. Patients responding poorly to treatment displayed a marginally lower microbiome profile distance from baseline, compared to those responding favourably. Conclusion and clinical relevance: Periodontal treatment does not alter the broader salivary microbiome profile, but may have selective implications on the species level. Treatment outcome can be impactful in the microbiome profile, as reduced changes may be associated with poorer responses. This article is protected by copyright. All rights reserved
Article
Full-text available
Dental caries and periodontitis account for a vast burden of morbidity and healthcare spending, yet their genetic basis remains largely uncharacterized. Here, we identify self-reported dental disease proxies which have similar underlying genetic contributions to clinical disease measures and then combine these in a genome-wide association study meta-analysis, identifying 47 novel and conditionally-independent risk loci for dental caries. We show that the heritability of dental caries is enriched for conserved genomic regions and partially overlapping with a range of complex traits including smoking, education, personality traits and metabolic measures. Using cardio-metabolic traits as an example in Mendelian randomization analysis, we estimate causal relationships and provide evidence suggesting that the processes contributing to dental caries may have undesirable downstream effects on health.
Article
Full-text available
Periodontitis is a microbial-induced chronic inflammatory disease, which may not only result in tooth loss, but can also contribute to the development of various systemic diseases. The transition from healthy to diseased periodontium depends on microbial dysbiosis and impaired host immune response. Although periodontitis is a common disease as well as associated with various systemic inflammatory conditions, the taxonomic profiling of the salivary microbiota in periodontitis and its association with host immune and inflammatory mediators has not been reported. Therefore, the aim of this study was to identify key pathogens and their potential interaction with the host's inflammatory mediators in saliva samples for periodontitis risk assessment. The microbial 16S rRNA gene sequencing and the levels of inflammatory mediators were performed in saliva samples from patients with chronic periodontitis and periodontally healthy control subjects. The salivary microbial community composition differed significantly between patients with chronic periodontitis and healthy controls. Our analyses identified a number of microbes, including bacteria assigned to Eubacterium saphenum, Tannerella forsythia, Filifactor alocis, Streptococcus mitis/parasanguinis, Parvimonas micra, Prevotella sp., Phocaeicola sp., and Fretibacterium sp. as more abundant in periodontitis, compared to healthy controls. In samples from healthy individuals, we identified Campylobacter concisus, and Veillonella sp. as more abundant. Integrative analysis of the microbiota and inflammatory mediators/cytokines revealed associations that included positive correlations between the pathogens Treponema sp. and Selenomas sp. and the cytokines chitinase 3-like 1, sIL-6Rα, sTNF-R1, and gp130/sIL-6Rβ. In addition, a negative correlation was identified between IL-10 and Filifactor alocis. Our results reveal distinct and disease-specific patterns of salivary microbial composition between patients with periodontitis and healthy controls, as well as significant correlations between microbiota and host-mediated inflammatory cytokines. The positive correlations between the pathogens Treponema sp. and Selenomas sp. and the cytokines chitinase 3-like 1, sIL-6Rα, sTNF-R1, and gp130/sIL-6Rβ might have the future potential to serve as a combined bacteria-host salivary biomarker panel for diagnosis of the chronic infectious disease periodontitis. However, further studies are required to determine the capacity of these microbes and inflammatory mediators as a salivary biomarker panel for periodontitis.
Article
Purpose: This study aimed to validate label-free quantitative proteomics (LFQ) against antibody-based methods for quantifying established periodontal disease biomarkers in saliva. Experimental design: In an experimental gingivitis model, healthy volunteers (n = 10) provided whole unstimulated saliva at baseline (d0), during the induction (d7, d14, d21) and resolution (d35) of gingival inflammation (total n = 50). Levels of four model biomarkers were analyzed by LFQ and time-resolved immunofluorometric assay (IFMA) or ELISA. Molecular matrix metalloproteinase (MMP)-8 forms were assessed by Western blot analysis. Results: LFQ detected elevated MMP-8 (d21 and d35 vs d7, p<0.05) and tissue inhibitor of matrix metalloproteinases (TIMP)-1 (d35 vs d7, p<0.05). Latent MMP-8 (70-80kDa) was present (d0-d35), but not active MMP-8 (50-60kDa). LFQ and immunoassay data correlated for MMP-8 (r = 0.36, p<0.001), myeloperoxidase (r = 0.39, p<0.0001), polymorphonuclear leukocyte elastase (r = 0.33, p<0.001) and TIMP-1 (r = -0.24, p<0.05). Conclusion and clinical relevance: LFQ can identify and quantify enzyme levels in saliva, however lack the ability to measure enzymatic activity. WB analysis revealed that MMP-8 may not be activated in the course of induction of gingival inflammation. Significant but weak correlations between IFMA or conventional ELISAs and LFQ suggest a limited capacity of available antibodies to reliably quantify salivary biomarkers for periodontal diseases. Novel 'anti-peptide' antibodies designed by newer targeted MS-based approaches could help to overcome these drawbacks. This article is protected by copyright. All rights reserved.
Article
Objective: To systematically assess the diagnostic value of host-derived salivary biomarkers based on their reported sensitivity and specificity in relation to clinical parameters of periodontal disease diagnosis in adults. Materials and methods: Comprehensive search of PubMed, Nature, Cochrane, OVID (Embase, MEDLINE [R] and PsycINFO) was conducted up to August, 1st 2018, using key terms relevant to the research questions and Cochrane methodology, supplemented by a grey literature search. The revised Quality Assessment of Diagnostic Accuracy Studies (QUADAS- 2) tool was used to assess the methodological quality of all included studies. Results: Seven studies were included in the review. Macrophage inflammatory protein-1αlpha (MIP-1α), interleukin-1beta (IL-1β), interleukin-6 (IL-6) and matrix metalloproteinase-8 (MMP-8) were identified as diagnostically acceptable biomarkers for periodontal disease. Overall, the combination of IL-6 and MMP-8 showed best diagnostic performance. Also, a combination of the four key biomarkers (IL-1β, IL-6, MMP-8 and MIP-1α) showed promising results for distinction between gingivitis and periodontitis, as well as for periodontitis compared with gingival health. Results are interpreted with caution due to limitations in the number of studies included and their quality. Conclusion: Certain salivary biomarkers can potentially be useful in combination and singularly for the diagnosis of periodontal disease. However, further methodically robust research is required to validate these biomarkers.
Chapter
Periodontal disease entails the inflammatory destruction of the tooth supporting (periodontal) tissues as a result of polymicrobial colonization of the tooth surface in the form of biofilms. Extensive data collected over the past decades on this chronic disease demonstrate that its progression is infrequent and episodic, and the susceptibility to it can vary among individuals. Physical assessments of previously occurring damage to periodontal tissues remain the cornerstone of detection and diagnosis, whereas traditionally used diagnostic procedures do neither identify susceptible individuals nor distinguish between disease-active and disease-inactive periodontal sites. Thus, more sensitive and accurate "measurable biological indicators" of periodontal diseases are needed in order to place diagnosis (e.g., the presence or stage) and management of the disease on a more rational less empirical basis. Contemporary "omics" technologies may help unlock the path to this quest. High throughput nucleic acid sequencing technologies have enabled us to examine the taxonomic distribution of microbial communities in oral health and disease, whereas proteomic technologies allowed us to decipher the molecular state of the host in disease, as well as the interactive cross-talk of the host with the microbiome. The newly established field of metaproteomics has enabled the identification of the repertoire of proteins that oral microorganisms use to compete or co-operate with each other. Vast such data is derived from oral biological fluids, including gingival crevicular fluid and saliva, which is progressively completed and catalogued as the analytical technologies and bioinformatics tools progressively advance. This chapter covers the current "omics"-derived knowledge on the microbiome, the host and their "interactome" with regard to periodontal diseases, and addresses challenges and opportunities ahead.