diagnostics

Article
Salivary Immune and Metabolic Marker Analysis
(SIMMA): A Diagnostic Test to Predict Caries Risk
Alex Mira *, Alejandro Artacho, Anny Camelo-Castillo, Sandra Garcia-Esteban and
Aurea Simon-Soro
Center for Advanced Research in Public Health, FISABIO Foundation, Valencia 46020, Spain;
artacho_ale@gva.es (A.A.); camelo_ann@gva.es (A.C.-C.); garcia_sanest@gva.es (S.G.-E.);
simon.aurea@gmail.com (A.S.-S.)
* Correspondence: mira_ale@gva.es; Tel.: +34-961-925-925

Received: 21 April 2017; Accepted: 16 June 2017; Published: 27 June 2017

Abstract: By using ELISA and colorimetric tests, we have measured 25 compounds in individuals
with and without dental caries at different time points of dental biofilm formation and time of
the day. We find that some compounds appear to be affected by circadian rhythms, others by
dental plaque maturity, and others show constant values during a 24 h period. Using univariate
analysis and cross-validation techniques, we have selected six components measured at specific time
points that maximize the diagnostic separation of health and disease conditions. Two out of the six
selected compounds are related to immune competence, another two to the adhesion capacity of
micro-organisms, and another two to acid production or pH buffering. We conclude that, in order to
design a robust caries risk test, the time of saliva sampling must be standardized and biomarkers
from different categories must be included. The preliminary data shown in this paper provide a
proof of principle of a caries risk test based on risk-associated categories. Thus, the test will provide
not only a general caries risk assessment, but also the likely biological origin of that risk, namely:
immune imbalance, and/or a tendency to adhesion of cariogenic organisms, and/or a lack of acid
buffering. When tested longitudinally and validated in larger cohorts, this could open the possibility
to develop preventive and personalized treatments.

Keywords: saliva; dental caries; circadian rhythms; immune system; buffering capacity; pH; adhesion;
microorganisms; toothpaste

1. Introduction
Dental caries (tooth decay) is the most prevalent chronic disease in the world. Data from the
World Health Organization indicate that 80% of the human population suffers or has suffered from
it, and it affects over 50% of the population at school age [1]. Dental caries is caused by the acid
produced by micro-organisms inhabiting the oral cavity, as a consequence of the fermentation of
dietary sugars. This lowers the pH on the tooth surface under a certain threshold, below which the
enamel demineralizes, initiating a caries lesion [2]. Once the lesion is cavitated, it is irreversible and the
damage can only be restored through clinical intervention, for instance a restoration or a tooth implant.
Although cavities are caused by micro-organisms, dental caries is a multi-factorial disease [3].
Apart from the microbiology, both human-related factors such as immune competence, enamel strength,
tooth shape, or saliva buffering effect, and external environmental factors such as diet, oral hygiene, or
fluoride exposure have a direct impact on tooth decay rates [4].
Despite its high prevalence and its direct and indirect impact on human health, there are still no
effective diagnostic tools to predict dental caries, and therefore dedicate the appropriate personalized
measures to prevent the disease. A large effort has been dedicated to study bacterial composition
in the oral cavity [59], with the aim of developing tests that could relate the presence of acidogenic

Diagnostics 2017, 7, 38; doi:10.3390/diagnostics7030038 www.mdpi.com/journal/diagnostics

Diagnostics 2017, 7, 38 2 of 14

organisms to caries risk. This has been performed by the direct culturing of specific bacteria [10], or by
DNA-based chips that would identify the presence of potentially pathogenic species [11]. The former
approach has been clinically applied for years, and includes diagnostic kits for caries risk assessment,
focused on culturing the acidogenic mutans streptococci and lactobacilli. However, the bacterial
counts of these two bacteria, which have traditionally been considered the main etiological agents of
dental caries, have proven to provide limited diagnostic value to predicting disease progression [12].
This can be due to the fact that Streptococcus mutans accounts for less than 1% of the total bacterial
community in cavities [17], and that lactobacilli are only found in dentin cavities and not in initial
enamel caries lesions, indicating that this species is not involved in caries initiation [13]. Furthermore,
the bacterial community within cavities is variable among individuals and cavities [14], hampering
the identification of a unique list of pathogenic organisms with predictive value for this polymicrobial
disease. In addition, most bacteria-based tests utilize saliva as a sample [9,15], and it has been shown
that saliva is not representative of the microbial community at the disease site [14,16,17].
Although caries risk tests based on saliva microbial composition have not been successful in
predicting tooth decay, saliva is a powerful sample for measuring overall body health. It is the
most easily available and accessible body fluid, and markers expressed in saliva can be used for the
diagnosis and patient follow-up of different diseases, including cancer, diabetes, hereditary disorders,
and infections [18]. Given that saliva contains many molecules that can directly or indirectly influence
oral micro-organisms and their potential cariogenic effect, measuring the levels of appropriate salivary
compounds may provide information to predict caries risk [19].
Saliva contains a wealth of substances that are protective of the teeth, and salivary flow has been
shown to be a key determinant of protection against tooth decay [20]. In fact, diseases with impaired
salivary function, as well as medication reducing saliva production, are frequently accompanied
by increased frequencies of tooth decay [21]. There are numerous salivary molecules that could
theoretically, and in many cases experimentally, be related to caries propensity, and most of them can
be included within three general categories:

1. Molecules related to the Immune System. These include immunoglobulins, antimicrobial

peptides, and proteins of the component system, which constitute a protection barrier against
oral pathogens [22].
2. Molecules related to the adhesion capacity of micro-organisms. These include structural
components of saliva that microorganisms use as targets for sticking to the tooth and forming the
dental plaque [23].
3. Molecules related to the acidity of saliva and plaque. These include enzymes that metabolize
sugars, acidic compounds that are produced as a consequence of sugar fermentation, and those
salivary components that can act as acid neutralizers [24].

A cautionary aspect to be considered when searching for potential caries-associated salivary

molecules is that there are several kinds of saliva collection protocols. These include unstimulated,
drooling saliva, as well as stimulated saliva after paraffin chewing gum, collection with paper points,
oral rinse with saline solution, collection with sterile swabs, or spitting [25], all of which will affect
the levels of the different compounds to be measured. In addition, the concentration of salivary
components will vary during the day as part of the normal circadian rhythms and changes in salivary
flow [26]. As a consequence, for the reliable measure of salivary molecules, a specific sampling protocol
and collection time are necessary.
The current work aims at identifying salivary molecules of the three kinds indicated above that
could vary in concentration between caries-free and caries-prone individuals, and that could be used as
biomarkers of caries risk. We have selected a list of 25 compounds belonging to these three categories
that are supported in the literature as potentially or theoretically linked to dental caries, and measured
them in caries-free and caries-active adults at four different times during a 24 h period in order to select
those molecules and time points with potential diagnostic value. In addition, to predict the caries
Diagnostics 2017, 7, 38 3 of 14

tendency of an individual, the identification of the underlying reason for the diseases propensity
could allow the design of person-specific preventive treatments in the future. Thus, this preliminary
work aims to identify the potential biomarkers of dental caries, and establish the approximate health-
and disease-associated concentrations of those compounds.

2. Materials and Methods

2.1. Donor Selection and Sampling Procedure

An intraoral examination was performed on 20 subjects, gathering information on the presence of
caries (including cavitated and non-cavitated enamel lesions as well as dentin caries lesions), plaque
deposits (Oral Hygiene Index (OHI)) [27], and the presence of gingival bleeding (Le and Silness
gingival index (GI)) [28], following the recommendations and nomenclature from the World Health
Organization [29]. They had not been treated with antibiotics in the three months prior to the study,
nor presented with the antecedents of the routine use of oral antiseptics. They were all adults aged
1939 years. Ten subjects had no history of dental caries (Decayed, Missing and Filled Teeth (DMFT)
index = 0), and the other ten had active cavities at the moment of sampling, with or without a history
of dental caries (fillings). All of the donors signed an informed consent, and the protocol was approved
by the Ethics Committee of the Valencian Public Health Authority (FISABIO-DGSP). The clinical
characteristics of subjects are indicated in Table S1.
All of the donors brushed their teeth at 910 a.m. using water, to prevent any potential effect
of toothpaste on salivary composition and characteristics. Five milliliters of non-stimulated saliva
samples were taken by drooling at 30 min, 6, 12, and 24 h after toothbrushing, collecting it in a sterile
50 mL Falcon tube while avoiding spitting or plaque removal by the tongue [30]. The samples were
immediately frozen at 80 C until used. For testing the effect of sugar, drooling saliva samples were
collected 10 min after a 1 min oral rinse with a 10% sugar solution.

2.2. Quantification of Salivary Immune Components

Enzyme-linked immunosorbent assays (ELISA) and colorimetric tests were performed by
duplicate to measure the salivary concentrations of 25 compounds, following the manufacturers
recommendations. A list of all 25 compounds, including the manufacturer information and the sample
dilution used, is shown in Table 1.

Table 1. Salivary components measured in the current work and manufacturer of the kit used for
determining their concentrations.

Category Salivary Compound Manufacturer Sample Dilution

IgA Assaypro 1:1000
IgG Assaypro 1:10
IgM Assaypro 1
Alpha-defensin 1-3 Assaypro 1:10
-defensin 1 Sun Red 1
-defensin 2 Sun Red 1
Immune System
-defensin 3 Sun Red 1
Cathelicidin LL-37 Assaypro 1:200
Lactoferrin Assaypro 1:20
Calprotectin Assaypro 1
Lysozyme Assaypro 1:8000
C3a Sun Red 1:10
Diagnostics 2017, 7, 38 4 of 14

Table 1. Cont.

Category Salivary Compound Manufacturer Sample Dilution

PRB1 Sun Red 1:10
Statherin Sun Red 1:200
Collagen type I Sun Red 1
Adhesion Mucin C7 Sun Red 1:50
Mucin C5B Sun Red 1:100
Alpha-2 macroglobulin Assaypro 1:4
Fibronectin Assaypro 1:20
Lactate BioVision 1:20
Formate BioVision 1:5
Calcium BioVision 1
pH
Phosphate BioVision 1:200
Urea BioVision 1:25
Alpha-amylase Biovision 1:10

2.3. Statistical Analysis

We focused on two requirements in terms of usability expectations and model robustness. First,
it should include at least two variables from each of the groups adhesion components, acid
production/buffering, and immune components. Second, it should be able to provide to the patient
two different and complementary risk measurements: on the one hand a univariant-obtained local risk
based on the comparison of the patients values of selected variables to confidence intervals calculated
on caries-free individuals, and on the other hand a multivariant-obtained global risk provided by
the overall model. Local-wise analysis would let us design modular therapies focused on one of the
groups adhesion, acid production/buffering, and immune, and global-wise analysis would give
us an emergency degree for treatment, in order to take those variables out of the confidence intervals
and back to normal values.
In order to select the variables included in our model, the Wilcoxon test as implemented in
an R environment [31] was performed on each of the candidate variables. The lower the p-value
provided by the test, the higher is the capability of a variable to distinguish between two groups
of samples (Caries-Free and Caries-Active). A non-parametric approach has been adopted to avoid
making assumptions about the variables distribution. Apart from exhibiting a significant p-value, a
requirement of a variable to be included in our model is that the confidence intervals (given by lower
and upper quantiles) corresponding to the groups of samples Caries and No-Caries do not overlap.
In order to assess the classification accuracy of the variables selected, the k-fold cross-validation
technique as implemented in a Galgo R package [32] has been adopted. The dataset was split into k
different training and test sets, and the classification accuracy was then defined as the average of the
classification accuracies of a model trained on training sets and calculated on the test sets for each
of the k splits. Currently, only the set of variables to be included in the model is known; nothing is
known about their interactions structure. For this reason, we adopted a single-hidden-layer neural
network model implemented in a nnet R package [33] and offered by Galgo as an unsupervised
approach to calculate the classification accuracy of the variables selected.

3. Results

3.1. Selection of Sampling Time

An initial test was performed with 10 compounds (IgA, IgG, IgM, -defensin 13, -defensin
1, -defensin 2, -defensin 3, LL-37, Lactoferrin, and Calprotectin), which were measured in saliva
samples taken at 0.5, 6, 12, and 24 h after toothbrushing. These sampling points corresponded to
910 a.m., 34 p.m., 910 p.m., and 910 a.m. the next morning, respectively. By using these four
moments, the potential effect of daily rhythms, as well as the effect of dental plaque maturity, could
Diagnostics 2017, 7, 38 5 of 14

be evaluated. A period of 30 min after toothbrushing was chosen, to allow for the stabilization of
the salivary concentrations that could be altered due to the mechanical tissue abrasion. Important
concentration changes were observed across time for most compounds, indicating that the salivary
levels of these proteins are not constant (Figure 1). The trends were, however, different depending
on the compound. IgG, for instance, showed a decrease in salivary concentration from the time of
toothbrushing, whereas Calprotectin displayed an increase through time. This suggests that the time
of toothbrushing could have an effect on the salivary concentrations of some compounds. For some
proteins, such as IgA, -defensin 2, or -defensin 3, a clear U-shape pattern was observed for caries-free
individuals, where the concentrations decreased during the afternoon and night but were higher in
Diagnostics 2017, 7, 38 5 of 13
the two morning samples, suggesting an influence of circadian daily rhythms (the p-values for the
comparison
individualsbetween
at anotherthetime
morning
point. and afternoon samples
We hypothesize that thiswere 0.019
can be one for IgA,
of the 0.0005why
reasons for -defensin
the results 2,
and 0.019 for -defensin 3; the p-values for the comparison between the
of salivary tests which do not specify a sampling time may lack accuracy or consistency. two morning samples were,
respectively, 0.11, 0.35, and 0.58 (Wilcox test)). Interestingly, the salivary concentrations
Given that 12 and 24 h after toothbrushing will not represent a comfortable and reliable sampling of IgA and
-defensin 2 in caries-active
time for clinical use, and thatindividuals
the sampling appeared to be
has ideally to constant
be adjusted through time,opening
to a clinics as a consequence
hours, the of
which the levels
morning of these two
and afternoon compounds
timepoints, in the afternoon
corresponding to 0.5and evening
and 6 h afterweretoothbrushing,
significantly different
were
considered for further study, and the measurements of all 25 salivary components
between the caries-active and caries-free groups (Table 2). Thus, the molecules that could be good were performed
at these two
biomarkers oftimepoints.
the diseaseThe at measured
a given time concentrations of the selectedbetween
may not discriminate 25 components
healthyfrom
andthe three
caries-risk
categories at
individuals foranother
caries-free
timeand caries-active
point. individuals
We hypothesize thatare indicated
this in Figure
can be one of theS1A, C and
reasons E (values
why at
the results
0.5 h, morning sample) and Figure S1B, D and F (values at 6 h, afternoon
of salivary tests which do not specify a sampling time may lack accuracy or consistency. sample).

Figure Temporalchanges
Figure1.1. Temporal changes in salivary
in salivary biomarkers.
biomarkers. The graphsTheshowgraphs
the show the concentrations
concentrations (means
(means
standard standard
error (SE))error (SE))
of 10 of 10 salivary
salivary immuneimmune
componentscomponents
in cariesinfree
caries
(n =free
10)(n = 10)
and and caries-active
caries-active (n =
(n 10)
= 10) individuals
individuals at four
at four time-points
time-points with with
a 24 ha period.
24 h period. Toothbrushing
Toothbrushing was performed
was performed at 9 a.m.atwith
9 a.m.
water.
with water.Samples
Samples were collected
were at 30
collected min,
at 30 6, 12,
min, and
6, 12, 2424
and hh after
aftertoothbrushing.
toothbrushing. Several
Severalcompounds
compounds
increaseorordecrease
increase decreaseininconcentration
concentration with with time
time after
after toothbrushing.
toothbrushing. Other Othersalivary
salivarycomponents
components
(marked
(marked witha adaynight
with daynightsymbol)
symbol) display
display a U-shape
U-shape pattern
patternwhere
wherethe thetwo
twomorning
morning samples
samples have
have
similar
similar concentrations,
concentrations, suggesting
suggesting that that
they they are influenced
are influenced by circadian
by circadian rhythms.rhythms.
Potential Potential
biomarkers
biomarkers
include LL37,include
which LL37,
appears which appears to discriminate
to discriminate between caries-free
between caries-free and caries-active
and caries-active groups at groups
all time
at all time points, or -defensin 2, which shows large differences between caries-free
points, or -defensin 2, which shows large differences between caries-free and caries-active individualsand caries-active
individuals
only only in the
in the afternoon andafternoon
evening and evening samples.
samples.
Diagnostics 2017, 7, 38 6 of 14

Table 2. Statistical significance for the comparison of 25 measured salivary compounds between
caries-free and caries-active individuals at two different times after tooth brushing (performed at 9 a.m.).

Category Compound 0.5 h 6h

Cathelicidin LL-37 0.007 0.002
IgA 0.007 0.353
-Defensin 1-3 0.105 0.853
IgM 0.326 1.000
-Defensin 1 0.393 0.796
Lactoferrin 0.393 1.000
Immune System
C3a 0.393 0.315
Calprotectin 0,405 0.326
-Defensin 3 0.520 0.970
IgG 0.520 0.121
-Defensin 2 0.971 7.57 105
Lysozyme 0.796 0.280
Statherin 0.063 0.247
Fibronectin 0.123 0.121
Mucin 5B 0.218 0.684
Adhesion Mucin 7 0.247 0.571
Collagen I 0.273 0.104
Alpha-2 Macroglobulin 0.393 0.796
PRB1 0.912 0.190
Phosphate 0.140 0.054
Lactate 0.353 0.123
pH Urea 0.393 0.393
Calcium 0.631 0.315
Formate 0.853 0.029

Given that 12 and 24 h after toothbrushing will not represent a comfortable and reliable sampling
time for clinical use, and that the sampling has ideally to be adjusted to a clinics opening hours, the
morning and afternoon timepoints, corresponding to 0.5 and 6 h after toothbrushing, were considered
for further study, and the measurements of all 25 salivary components were performed at these two
timepoints. The measured concentrations of the selected 25 components from the three categories
for caries-free and caries-active individuals are indicated in Figure S1A,C,E (values at 0.5 h, morning
sample) and Figure S1B,D,F (values at 6 h, afternoon sample).

3.2. Selection of Caries-Associated Biomarkers

The medians and upper/lower quartiles of all of the salivary components in caries-free and
caries-active individuals are shown for the 0.5 h measurements (Figure S1A,C,E and for the 6 h
measurements (Figure S1B,D,F) for immune molecules, adhesion molecules, and acid/buffering
components. Wilcox univariate tests were performed to compare the values between individuals with
and without caries (Table 2). As can be observed, few of the measured variables in fact have diagnostic
value, even if they belong to the same category. The medians and interquartile ranges of the two
components from each category with the best discriminating capacity are shown in Figure 2.
Diagnostics 2017, 7, 38 7 of 14
Diagnostics 2017, 7, 38 7 of 13

Figure
Figure 2. Concentrationsofofpotential
2. Concentrations potentialsalivary
salivarybiomarkers
biomarkers of dental
dental caries.
caries. The
Theboxplots
boxplotsrepresent
represent
normalized median and interquartile ranges for 10 caries-active (CA) and 10 caries-freecaries-free
normalized median and interquartile ranges for 10 caries-active (CA) and 10 (CF)
(CF) individuals
in 6individuals in 6 salivarywith
salivary components components
potentialwith potentialvalue
diagnostic diagnostic
at two value at two time-points:
time-points: (A) 0.5
(A) 0.5 h after h
tooth
after tooth
brushing brushing
(morning (morning
sample); andsample); and (B)
(B) 6 h after 6 hbrushing
tooth after tooth(afternoon
brushing (afternoon
sample. sample.

The data show that immune components are the ones that better discriminate between healthy
and diseased individuals. This suggests an important role for immune competence in the risk of
Diagnostics 2017, 7, 38 8 of 13

Thus, based on the p-values from the univariate analyses (Table 2) and the lack of overlap
Diagnostics the 7,
between 2017, 38
data 8 of 14
dispersion boxes (Figure S1), the following salivary metabolites are selected to
provide discrimination value between healthy and caries-active individuals:
developing
1. At 6 h: caries. At 6 h after tooth brushing (afternoon sample), several components of each of the
three categories were different between the two patient groups, whereas at 0.5 h (morning sample)
a. Immune molecules: -defensin 2 and LL-37.
no differences were found in the acidic component category. The latter could be due to the fact that
b. Adhesion molecules: Collagen I and Fibronectin.
these metabolites are mainly produced after dietary carbohydrate fermentation, and are more readily
c. pH components: Formate and Phosphate.
measured at 6 h (after lunch in our sampling schedule). In order to test this possibility, the same test
was
2. repeated
At 0.5 h: in the morning but 10 min after a 1 min rinse with a 10% sugar solution. The results
show an improvement in the discriminatory power of Statherin and of compounds in the pH buffering
a. Immune molecules: LL-37 and IgA.
category, specifically Formate and Phosphate (Figure 3). Curiously, Lactates discriminatory power
b. Adhesion molecules: Statherin and Fibronectin (Statherin only if saliva is collected after a
did not improve after the sugar rinse. The biomarker concentrations in the other two categories were
sugary solution rinse).
affected by the sugar rinse, and although the overall tendency for the selected biomarkers in the
c. pH components: Phosphate and Lactate (Phosphate and Formate if saliva is collected after
adhesion and immune categories was maintained, the difference between caries-active and caries-free
a sugary solution rinse).
individuals was significant only for Statherin.

Figure 3. Effect
Figure 3. Effect of
of sugar
sugar rinse
rinse on
on salivary
salivary concentrations
concentrations ofof potential
potential biomarkers.
biomarkers. Unstimulated
Unstimulated saliva
saliva
samples were collected 10 min after a 1 min rinse with a 10% sucrose solution performed
samples were collected 10 min after a 1 min rinse with a 10% sucrose solution performed between between9
910 a.m.Relative
10 a.m. Relativetotothe
themorning
morningsamples
samplesininTable
Table22 and
and Figure
Figure 2,
2, Phosphate
Phosphate and
and Formate
Formate increase
increase
their
their discriminating
discriminating power.
power. The
The boxplots
boxplots show
show the
the median
median andand interquartile
interquartilerange
rangevalues.
values. Asterisks
Asterisks
indicate statistical significance (Wilcox rank sum test).
indicate statistical significance (Wilcox rank sum test).

In order
Thus, to on
based testthe
whether
p-valuesany
fromcombination of analyses
the univariate concentrations ofand
(Table 2) anythe
of lack
the aforementioned 25
of overlap between
compounds
the present boxes
data dispersion in saliva may improve
(Figure caries risk salivary
S1), the following prediction, the statistical
metabolites classification
are selected power
to provide
was comparedvalue
discrimination between the six
between variables
healthy selected above
and caries-active (those with the best p-values) and 1000
individuals:
random selections of variables. Power may be defined as (proportion of correct classification of caries
1. At 6 h: (CA)) + (proportion of correct classification of caries-free individuals (NOCA)). Thus, the
individuals
maximum classification power value is 2. When the potential biomarkers of caries risk were
a. Immune molecules: -defensin 2 and LL-37.
randomly selected in groups of six, the combinations did not improve the diagnostic value provided
b. Adhesion molecules: Collagen I and Fibronectin.
by the six selected compounds, neither at 0.5 h after brushing teeth (Figure 4A) or at 6 h (Figure 4B).
c. pH components: Formate and Phosphate.
Diagnostics 2017, 7, 38 9 of 14

2. At 0.5 h:

a. Immune molecules: LL-37 and IgA.

b. Adhesion molecules: Statherin and Fibronectin (Statherin only if saliva is collected after a
sugary solution rinse).
c. pH components: Phosphate and Lactate (Phosphate and Formate if saliva is collected after
a sugary solution rinse).

In order to test whether any combination of concentrations of any of the aforementioned

25 compounds present in saliva may improve caries risk prediction, the statistical classification
power was compared between the six variables selected above (those with the best p-values) and
1000 random selections of variables. Power may be defined as (proportion of correct classification
of caries individuals (CA)) + (proportion of correct classification of caries-free individuals (NOCA)).
Thus, the maximum classification power value is 2. When the potential biomarkers of caries risk were
Diagnostics 2017,
randomly 7, 38 in groups of six, the combinations did not improve the diagnostic value provided
selected 9 of 13
by the six selected compounds, neither at 0.5 h after brushing teeth (Figure 4A) or at 6 h (Figure 4B).
Specifically, the
Specifically, the median
median classification
classification power
power ofof the
the six
six randomly
randomly selected
selected compounds
compounds waswas 1.2
1.2 at
at both
both
0.5 and 6 h. Thus, it may be concluded that the six selected variables are those
0.5 and 6 h. Thus, it may be concluded that the six selected variables are those that maximize thethat maximize the
diagnostic value
diagnostic value of of all
all of
of the
the measured
measured biomarkers,
biomarkers, especially
especially in
in the
the afternoon
afternoon samples.
samples.

Figure 4.4.Statistical classification

Statistical power
classification of salivary
power components.
of salivary The X-axis
components. TheinX-axis
the frequency
in the histograms
frequency
indicate
histograms indicate the classification power of 1000 combinations of 6 components randomlyfrom
the classification power of 1000 combinations of 6 components randomly selected the
selected
25 compounds measured
from the 25 compounds measured in saliva in the current manuscript, including the 6 components
saliva in the current manuscript, including the 6 components selected
as the best
selected biomarkers
as the in the present
best biomarkers work, for
in the present samples
work, taken taken
for samples at (A) at
0.5(A)
h 0.5
(morning sample)
h (morning and
sample)
(B)
and6(B) h (afternoon
6 h (afternoonsample) afterafter
sample) brushing
brushingteeth.
teeth.Values
ValuesininthetheY-axis
Y-axisrepresent
represent thethe number
number of
random combinations for
random combinations for each
eachstatistical
statisticalpower
powercategory.
category.TheThe median
median classification
classification power
power of the
of the six
six
randomly selected compounds was 1.2 at both 0.5 and 6 h (marked with asterisks), whereas the
randomly selected compounds was 1.2 at both 0.5 and 6 h (marked with asterisks), whereas the
median
median classification
classificationpower
powerofofthetheselected
selectedbiomarkers
biomarkers (marked
(marked with redred
with vertical lines)
vertical waswas
lines) 1.6 1.6
andand2.0
at
2.00.5
at and 6 h after
0.5 and toothtooth
6 h after brushing, respectively.
brushing, PowerPower
respectively. = (proportion of correct
= (proportion classification
of correct of caries
classification of
individuals (CA)) + (proportion of correct classification of caries-free individuals
caries individuals (CA)) + (proportion of correct classification of caries-free individuals (NOCA)),(NOCA)), wherein
said proportions
wherein were estimated
said proportions using a 50-fold
were estimated using across-validation approach.approach.
50-fold cross-validation

selected variables was measured by aa cross-validation

The classification accuracy of the selected cross-validation
unsupervised approach [33], indicating that, on average, 98% of the caries individuals
individuals are detected
detected
both timepoints.
by the test at both timepoints. Increasing the number of variables from the six selected above to the
eight most significantly different compounds did not not improve
improve this
this percentage.
percentage. The cross-validation
cross-validation
technique using
technique usingthe
theGalgo
Galgo method
method for data
for data sampled
sampled at 6 hat 6 h afforded
afforded a sensitivity
a sensitivity of a98%
of 98% and and a
specificity
specificity
of of 88%.
88%. In other In other
words, words,
almost almost
100% 100% ofwith
of subjects subjects with
caries are caries are classified
classified correctly, correctly, whereas
whereas only 12%
only 12% of subjects without caries are falsely assigned to the high caries risk group. It is possible
that the 12% of false positives arise within the group of subjects without caries because these subjects
in fact have high caries risk, but have not clinically developed this condition due to, for example, the
quality of their diet and/or oral hygiene. Unfortunately, we did not collect diet data, and the validity
of this hypothesis should be tested with larger sample sizes, especially in longitudinal studies.
Diagnostics 2017, 7, 38 10 of 14

of subjects without caries are falsely assigned to the high caries risk group. It is possible that the 12%
of false positives arise within the group of subjects without caries because these subjects in fact have
high caries risk, but have not clinically developed this condition due to, for example, the quality of
their diet and/or oral hygiene. Unfortunately, we did not collect diet data, and the validity of this
hypothesis should be tested with larger sample sizes, especially in longitudinal studies.

4. Discussion
Based on the above preliminary data, a Salivary Immune and Metabolic Marker Analysis test
(SIMMA test) is proposed, which is based on measuring the salivary values from an individual at a
given time point of six selected compounds, two of which are related to immune competence, another
two to the adhesion capacity of micro-organisms, and another two to the acid production and buffering
capacity. Those values are then compared to the reference values obtained from a healthy population
of a similar age, and the concentrations falling outside the healthy range are indicative of caries risk
due to an imbalance in the corresponding category. Thus, the test will provide not only a general caries
risk assessment, but also the likely biological origin of that risk, namely: immune imbalance, and/or a
tendency to adhesion of cariogenic organisms, and/or a lack of acid buffering. Based on the SIMMA
test outcome, a preventive, personalized treatment will be possible, directed towards one 10
Diagnostics 2017, 7, 38
orofmore
13
of
the following goals: (i) immune modulation to select a non-cariogenic oral biofilm, which could be
towards
achieved, forone or morebyofprobiotic
example, the following goals:that
bacteria (i) immune
have beenmodulation
shown totostimulate
select a non-cariogenic oral
antibody production
(see [34] for a recent review); (ii) diminishing the adhesion capacity of a cariogenic biofilm,to
biofilm, which could be achieved, for example, by probiotic bacteria that have been shown which
couldstimulate
be achievedantibody production
by specific (see [34] for a molecules
anti-adherent recent review);
(see(ii)
fordiminishing the adhesion
example [35]) added tocapacity
daily of
dental
a cariogenic biofilm, which could be achieved by specific anti-adherent molecules (see for example
hygiene products; and (iii) improving buffering capacity, which could be achieved by stimulating
[35]) added to daily dental hygiene products; and (iii) improving buffering capacity, which could be
salivary flow through chewing or by the addition of buffering molecules or prebiotic compounds that
achieved by stimulating salivary flow through chewing or by the addition of buffering molecules or
stimulate ammonia
prebiotic compounds production (see for
that stimulate example
ammonia [36]) to daily
production dental
(see for hygiene
example products.
[36]) to A flow
daily dental chart of
hygiene
the SIMMA
products. test, its rationale,
A flow chart of theand applications
SIMMA is shown
test, its rationale, andinapplications
Figure 5. is shown in Figure 5.

Figure
Figure 5. 5.Rationale
Rationale of
of the
the Salivary
SalivaryImmune
Immune andand
Metabolic Marker
Metabolic Analysis
Marker (SIMMA)
Analysis test. An test.
(SIMMA)
unstimulated saliva sample is used to measure different compounds belonging to three functional
An unstimulated saliva sample is used to measure different compounds belonging to three functional
categories, and compare their concentrations to those or healthy, caries-free individuals from the same
categories, and compare their concentrations to those or healthy, caries-free individuals from the same
age. A skewed concentration for any of those biomarkers is considered to represent an imbalance in
age. A skewed concentration for any of those biomarkers is considered to represent an imbalance in
the corresponding category, opening possibilities for individual-specific preventive measures.
the corresponding category, opening possibilities for individual-specific preventive measures.
Our data also underline the importance of standardizing sampling time, because some
molecules with potential diagnostic value are subject to daily rhythms. Although we did not measure
salivary flow in our samples, observed daily changes in the levels of some compounds must partly
be due to salivary flow, which is known to follow a circadian rhythm [26], where lower saliva levels
in the morning would tend to elevate solute concentrations. Thus, if sampling time is not taken into
account, an individual salivary biomarker may have more predictive power when the data are
Diagnostics 2017, 7, 38 11 of 14

Our data also underline the importance of standardizing sampling time, because some molecules
with potential diagnostic value are subject to daily rhythms. Although we did not measure salivary
flow in our samples, observed daily changes in the levels of some compounds must partly be due
to salivary flow, which is known to follow a circadian rhythm [26], where lower saliva levels in the
morning would tend to elevate solute concentrations. Thus, if sampling time is not taken into account,
an individual salivary biomarker may have more predictive power when the data are normalized with
salivary flow rates or total protein concentration (a flow rate dependent parameter).
The univariate analysis determines the individual salivary compounds that, once measured
and compared to the healthy reference values, will suggest the appropriate treatment to prevent the
appearance of caries. In addition to this, it must be kept in mind that the combination of measurements
will be more informative and sensitive than individual ones. For instance, an individual may present
normal values for a given compound but have out-of-range values for another. This is one of the
reasons why tests based on individual variables will likely lack the sensitivity to detect the risk of
caries. In addition, not only the values of each compound but also the interaction among them may
provide information about an individuals caries risk. Thus, combining the values of all of the selected
compounds measured in a multivariate analysis should also be performed to provide an overall caries
risk value. This overall value will inform the clinician about the general tendency of the patient to
develop caries. In practice, the number of out-of-range compounds could also serve in the clinic as a
measure of the caries risk in a patient, and therefore their treatments urgency. This information could
also serve to determine individuals at risk, where the frequency of visits and the type of interventions
can be adapted to reduce the probability of future caries development [37,38].

5. Conclusions
In conclusion, a test based on the selection of biomarkers from different risk-associated categories
will provide an overall caries risk value and a list of salivary components that show skewed values.
The test should be performed at a specific timepoint and time since toothbrushing, given that both
factors, especially daily rhythms, affect salivary compounds concentration. If urine or blood tests
have to be performed under specific conditions or at timepoints for determining the health boundaries
of biomarkers, it is not unreasonable to assume that the same standardization has to be achieved with
salivary tests. The functional category to which those skewed components belong may provide a
putative prevention treatment to restore values to the healthy range.
A limitation of the current study is clearly the small sample size. Nevertheless, the preliminary
data shown in this paper provide a proof of principle of a caries risk test based on risk-associated
categories. The specific boundaries of health and disease in the concentrations of the different
biomarkers are likely to be age-specific, and should be quantified in study groups of different ages,
especially children, which is the group in which preventive strategies are most fruitful. Although
not shown in this paper, we measured salivary pH and pH buffering capacity in the same samples
using commercial kits, but these basic measurements failed to discriminate between caries-free and
caries-active individuals. However, we did not measure other variables normally used for caries
risk assessment, such as the salivary levels of cariogenic organisms, or dietary habits. Thus, the
test proposed in this paper should be compared with the methods that are currently accepted in
the assessment of caries risk (see, for example, [37,38]). Once the appropriate biomarkers have been
selected, the SIMMA test should be transformed from the current laboratory measurements into a
ready-to-use kit based on reactive strips, where out-of-range values for one or two biomarkers per
category can be easily and quickly visualized without the need for laboratory equipment. The use
of diagnostic strips has been successfully applied to determine the risk of periodontal disease in
adolescents based on the levels of the human matrix metalloproteinase MMP-8 [39]. Similarly, the
development of point-of-care diagnostic strips could be instrumental for an application of caries risk
tests at a community level. Especially relevant would be the application of caries risk assessment
in children, in order to determine those individuals at high risk where preventive measures could
Diagnostics 2017, 7, 38 12 of 14

be implemented. Some of those, like the sealing of pits and fissures for caries prevention, would
be too costly and unnecessary to perform on all children, and a test able to select high-risk patients
would be extremely helpful [40]. In private clinical practice, the identification of high-risk individuals,
and especially the putative cause of the risk, would provide the dentist with valuable information
to personalize the treatment, as well as to establish the timing of visits. The development of caries
risk assessment methods in order to achieve personalized, precision dentistry is both desirable
and achievable, but several conceptual and analytical mistakes have been highlighted, including the
application of population-level variables to individuals or the use of inappropriate modeling [41].
The data presented in this paper show an association of some salivary components with an existing
caries status. When appropriate health thresholds are established for the different biomarkers,
longitudinal studies will determine whether those compounds are not only disease-associated, but
have also a predictive value.

Supplementary Materials: The following are available online at www.mdpi.com/2075-4418/07/3/38/s1.

Acknowledgments: This work was supported by grant FISABIO UGP-15-225. We did not receive funds for
covering the costs to publish in open access.
Author Contributions: Alex Mira and Aurea Simon-Soro conceived and designed the experiments;
Aurea Simon-Soro, Anny Camelo-Castillo and Sandra Garcia-Esteban and performed the experiments;
Aurea Simon-Soro, Alex Mira and Alejandro Artacho analyzed the data; Alex Mira contributed reagents and
materials; Alex Mira wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest. The public founding sponsors had no role in the
design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, and in
the decision to publish the results.

References
1. Petersen, P.E. Challenges to improvement of oral health in the 21st centuryThe approach of the WHO
Global Oral Health Programme. Int. Dent. J. 2004, 54, 329343. [CrossRef] [PubMed]
2. Takahashi, N.; Nyvad, B. The Role of Bacteria in the Caries Process: Ecological Perspectives. J. Dent. Res.
2011, 90, 294303. [CrossRef] [PubMed]
3. Keyes, P.H.; Jordan, H.V. Factors influencing initiation, transmission and inhibition of dental caries.
In Mechanisms of Hard Tissue Destruction; Harris, R.J., Ed.; Academic Press: New York, NY, USA, 1963;
pp. 261283.
4. Ditmyer, M.M.; Dounis, G.; Howard, K.M.; Mobley, C.; Cappelli, D. Validation of a multifactorial risk factor
model used for predicting future caries risk with Nevada adolescents. BMC Oral Health 2011, 20, 1118.
[CrossRef] [PubMed]
5. Aas, J.A.; Paster, B.J.; Stokes, L.N.; Olsen, I.; Dewhirst, F.E. Defining the normal bacterial flora of the oral
cavity. J. Clin. Microbiol. 2005, 43, 57215732. [CrossRef] [PubMed]
6. Aas, J.A.; Griffen, A.L.; Dardis, S.R.; Lee, A.M.; Olsen, I.; Dewhirst, F.E.; Leys, E.J.; Paster, B.J. Bacteria of
dental caries in primary and permanent teeth in children and young adults. J. Clin. Microbiol. 2008, 46,
14071417. [CrossRef] [PubMed]
7. Crielaard, W.; Zaura, E.; Schuller, A.A.; Huse, S.M.; Montijn, R.C.; Keijser, B.J. Exploring the oral
microbiota of children at various developmental stages of their dentition in the relation to their oral health.
BMC Med. Genom. 2011, 4, 22. [CrossRef] [PubMed]
8. Belda-Ferre, P.; Alcaraz, L.D.; Cabrera-Rubio, R.; Romero, H.; Simn-Soro, A.; Pignatelli, M.; Mira, A. The oral
metagenome in health and disease. ISME J. 2012, 6, 4656. [CrossRef] [PubMed]
9. Simn-Soro, A.; Toms, I.; Cabrera-Rubio, R.; Catalan, M.D.; Nyvad, B.; Mira, A. Microbial geography of the
oral cavity. J. Dent. Res. 2013, 92, 616621. [CrossRef] [PubMed]
10. Plonka, K.A.; Pukallus, M.L.; Barnett, A.G.; Walsh, L.J.; Holcombe, T.H.; Seow, W.K. Mutans streptococci and
lactobacilli colonization in predentate children from the neonatal period to seven months of age. Caries Res.
2012, 46, 213220. [CrossRef] [PubMed]
Diagnostics 2017, 7, 38 13 of 14

11. Olson, J.C.; Cuff, C.F.; Lukomski, S.; Lukomska, E.; Canizales, Y.; Wu, B.; Crout, R.J.; Thomas, J.G.;
McNeil, D.W.; Weyant, R.J.; et al. Use of 16S ribosomal RNA gene analyses to characterize the bacterial
signature associated with poor oral health in West Virginia. BMC Oral Health 2011, 11, 7. [CrossRef] [PubMed]
12. Tellez, M.; Gray, S.L.; Gray, S.; Lim, S.; Ismail, A.I. Sealants and dental caries: Dentists perspectives on
evidence-based recommendations. J. Am. Dent. Assoc. 2011, 142, 10331040. [CrossRef] [PubMed]
13. Shah, A.G.; Shetty, P.C.; Ramachandra, C.S.; Bhat, N.S.; Laxmikanth, S.M. In vitro assessment of
photocatalytic titanium oxide surface modified stainless steel orthodontic brackets for antiadherent and
antibacterial properties against Lactobacillus acidophilus. Angle Orthod. 2011, 81, 10281035. [CrossRef]
[PubMed]
14. Simn-Soro, A.; Belda-Ferre, P.; Cabrera-Rubio, R.; Alcaraz, L.D.; Mira, A. A tissue-dependent hypothesis of
dental caries. Caries Res. 2013, 47, 591600. [CrossRef] [PubMed]
15. Yoshizawa, J.M.; Schafer, C.A.; Schafer, J.J.; Farrell, J.J.; Paster, B.J.; Wong, D.T. Salivary biomarkers: Toward
future clinical and diagnostic utilities. Clin. Microbiol. Rev. 2013, 26, 781791. [CrossRef] [PubMed]
16. Huang, S.; Yang, F.; Zeng, X.; Chen, J.; Li, R.; Wen, T.; Li, C.; Wei, W.; Liu, J.; Chen, L.; et al. Preliminary
characterization of the oral microbiota of Chinese adults with and without gingivitis. BMC Oral Health 2011,
11, 33. [CrossRef] [PubMed]
17. Simn-Soro, A.; Guillen-Navarro, M.; Mira, A. Metatranscriptomics reveals overall active bacterial
composition in caries lesions. J. Oral Microbiol. 2014, 6, 25443. [CrossRef] [PubMed]
18. Schafer, C.A.; Schafer, J.J.; Yakob, M.; Lima, P.; Camargo, P.; Wong, D.T. Saliva diagnostics: Utilizing oral
fluids to determine health status. Monogr. Oral Sci. 2014, 24, 8898. [PubMed]
19. Guo, L.; Shi, W. Salivary biomarkers for caries risk assessment. J. Calif. Dent. Assoc. 2013, 41, 107109,
112118. [PubMed]
20. Gao, X.; Jiang, S.; Koh, D.; Hsu, C.Y. Salivary biomarkers for dental caries. Periodontol. 2000 2016, 70, 128141.
[CrossRef] [PubMed]
21. Diaz de Guillory, C.; Schoolfield, J.D.; Johnson, D.; Yeh, C.K.; Chen, S.; Cappelli, D.P.; Bober-Moken, I.G.;
Dang, H. Co-relationships between glandular salivary flow rates and dental caries. Gerodontology 2014, 31,
210219. [CrossRef] [PubMed]
22. Tao, R.; Jurevic, R.J.; Coulton, K.K.; Tsutsui, M.T.; Roberts, M.C.; Kimball, J.R.; Wells, N.; Berndt, J.;
Dale, B.A. Salivary antimicrobial peptide expression and dental caries experience in children.
Antimicrob. Agents Chemother. 2005, 49, 38833888. [CrossRef] [PubMed]
23. Jenkinson, H.F.; Lamont, R.J. Oral microbial communities in sickness and in health. Trends Microbiol. 2005,
13, 589595. [CrossRef] [PubMed]
24. Loke, C.; Lee, J.; Sander, S.; Mei, L.; Farella, M. Factors affecting intra-oral pHA review. J. Oral Rehabil.
2016, 43, 778785. [CrossRef] [PubMed]
25. Khurshid, Z.; Zohaib, S.; Najeeb, S.; Zafar, M.S.; Slowey, P.D.; Almas, K. Human Saliva Collection Devices for
Proteomics: An Update. Int. J. Mol. Sci. 2016, 17, 846. [CrossRef] [PubMed]
26. Dawes, C. Circadian rhythms in human salivary flow rate and composition. J. Physiol. 1972, 220, 529545.
[CrossRef] [PubMed]
27. Greene, J.C.; Vermillion, J.R. The Simplified Oral Hygiene Index. J. Am. Dent. Assoc. 1964, 68, 713. [CrossRef]
[PubMed]
28. Lo, H.; Silness, J. Periodontal disease in pregnancy. Acta Odontol. Scand. 1963, 21, 533. [CrossRef] [PubMed]
29. World Health Organization. Oral Health SurveysBasic Methods, 4th ed.; World Health Organization: Geneva,
Switzerland, 1997.
30. Navazesh, M.; Christensen, C.M. A comparison of whole mouth resting and stimulated salivary measurement
procedures. J. Dent. Res. 1982, 61, 11581162. [CrossRef] [PubMed]
31. R Core Team. R: A Language and Environment for Statistical Computing; R Foundation for Statistical Computing:
Vienna, Austria, 2014; Available online: http://www.R-project.org/ (accessed on 7 October 2015).
32. Trevino, V.; Falciani, F. Galgo: Genetic Algorithms for Multivariate Statistical Models from Large-scale
Functional Genomics Data. R Package Version 1.2. 2014. Available online: http://bioinformatica.mty.itesm.
mx/?q=node/82 (accessed on 7 October 2015).
33. Venables, W.N.; Ripley, B.D. Modern Applied Statistics with S, 4th ed.; Springer: New York, NY, USA, 2002.
34. Ashraf, R.; Shah, N.P. Immune system stimulation by probiotic microorganisms. Crit. Rev. Food Sci. Nutr.
2014, 54, 938956. [CrossRef] [PubMed]
Diagnostics 2017, 7, 38 14 of 14

35. Guan, Y.H.; Lath, D.L.; Graaf, T.; Lilley, T.H.; Brook, A.H. Moderation of oral bacterial adhesion on
saliva-coated hydroxyapatite by polyaspartate. J. Appl. Microbiol. 2003, 94, 456461. [CrossRef] [PubMed]
36. Santarpia, R.P.; Lavender, S.; Gittins, E.; Vandeven, M.; Cummins, D.; Sullivan, R. A 12-week clinical study
assessing the clinical effects on plaque metabolism of a dentifrice containing 1.5% arginine, an insoluble
calcium compound and 1450 ppm fluoride. Am. J. Dent. 2014, 27, 100105. [PubMed]
37. Featherstone, J.D. The caries balance: The basis for caries management by risk assessment. Oral Health
Prev. Dent. 2004, 2, 259264. [PubMed]
38. Young, D.A.; Featherstone, J.D. Implementing caries risk assessment and clinical interventions. Dent. Clin.
N. Am. 2010, 54, 495505. [CrossRef] [PubMed]
39. Heikkinen, A.M.; Nwhator, S.O.; Rathnayake, N.; Mntyl, P.; Vatanen, P.; Sorsa, T. Pilot Study on Oral
Health Status as Assessed by an Active Matrix Metalloproteinase-8 Chairside Mouthrinse Test in Adolescents.
J. Periodontol. 2016, 87, 3640. [CrossRef] [PubMed]
40. Crall, J.J.; Donly, K.J. Dental sealants guidelines development: 20022014. Pediatr. Dent. 2015, 37, 111115.
[PubMed]
41. Divaris, K. Predicting Dental Caries Outcomes in Children: A Risky Concept. J. Dent. Res. 2016, 95,
248254. [CrossRef] [PubMed]

2017 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 (http://creativecommons.org/licenses/by/4.0/).