**Electrical Monitoring as a Novel Route to Understanding the Aging Mechanisms of Carbon Nanotube-Doped Adhesive Film Joints**


Received: 19 March 2020; Accepted: 3 April 2020; Published: 8 April 2020

#### **Featured Application: Structural Health Monitoring of novel carbon nanotube doped adhesive joints under aging conditions.**

**Abstract:** Carbon fiber-reinforced plastic bonded joints with novel carbon nanotube (CNT) adhesive films were manufactured and tested under different aging conditions by varying the surfactant content added to enhance CNT dispersion. Single lap shear (SLS) tests were conducted in their initial state and after 1 and 2 months immersed in distilled water at 60 ◦C. In addition, their electrical response was measured in terms of the electrical resistance change through thickness. The lap shear strength showed an initial decrease due to plasticization of weak hydrogen bonds, and then a partial recovery due to secondary crosslinking. This plasticization effect was confirmed by differential scanning calorimetry analysis with a decrease in the glass transition temperature. The electrical response varied with aging conditions, showing a higher plasticity region in the 1-month SLS joints, and a sharper increase in the case of the non-aged and 2-month-aged samples; these changes were more prevalent with increasing surfactant content. By adjusting the measured electrical data to simple theoretical calculations, it was possible to establish the first estimation of damage accumulation, which was higher in the case of non-aged and 2-month-aged samples, due to the presence of more prevalent brittle mechanisms for the CNT-doped joints.

**Keywords:** carbon nanotubes; aging; structural health monitoring; water uptake; adhesive film; surfactant

#### **1. Introduction**

The increasing requirements of industry in terms of structural components make the development of novel materials necessary. In this context, composite structures present many advantages over conventional metallic alloys due to their exceptional specific properties that lead to energy efficiency and weight savings.

Therefore, the assembly of several composite parts is a challenging subject as the complexity of these structures is continuously increasing. For these purposes, bonded joints have some advantages over bolted ones as they are lightweight and avoid stress concentrations around the holes [1]. However, the inspection of adhesive joints sometimes is not always straightforward, since it involves

many complex techniques, such as fiber Bragg grating sensors or Lamb waves, which often do not give a complete overview of the quality of the bonded joint [2–4]. Therefore, it is necessary to develop novel procedures that do not involve complex data analysis techniques and are not detrimental to the physical properties of the joint.

In this regard, carbon nanotubes (CNTs) seem to be a very promising solution. Their exceptional properties [5–7] and the enhancement of the electrical conductivity that they induce when added to an insulator resin [8–11] makes them very useful for multifunctional applications [12,13]. In fact, their use in structural health monitoring (SHM) applications is now of interest because of their piezoresistive and tunneling properties that lead to high sensitivities [14–17].

The aim of this work is to exploit the superb physical properties of CNTs in developing novel multifunctional adhesives with an inherent self-sensing capability. To date, most research into reinforced bonded joints has been focused on paste adhesives [18–21]. They exhibit excellent sensing properties and are capable of properly monitoring strain and debonding [22–24]. These paste adhesives can be treated as nanoreinforced composites, with the CNT dispersion procedure representing a challenging subject that often involves complex and expensive techniques, such as three-roll milling [25–27]. Therefore, this work is focused on the effect that CNT addition has on adhesive films, which allows for better thickness control and is used for structural applications in the aircraft industry.

In previous studies, CNT reinforced adhesive films have demonstrated high sensing properties and a good capability to properly monitor crack evolution [28–30]. The dispersion procedure has also been optimized in order to achieve a degree of good homogenization without any substantial detriment to the mechanical properties [31]. This is achieved by means of ultrasonication of a CNT dispersion in an aqueous solution, which is assisted by the addition of a surfactant, namely, sodium dodecyl sulfate (SDS). The addition of SDS improves the mechanical dispersion of the CNTs in the aqueous solution [32–35].

The effect of CNT dispersion on the mechanical and electrical properties of carbon fiber reinforced plastic (CFRP) bonded joints in their initial state has been characterized in previous works [31]. It has been concluded that these CNT adhesive films do not induce a detrimental effect on mechanical performance and they have proved to have excellent monitoring capabilities by means of electrical measurements [28]. This work takes a further step by analyzing the potential and applicability of these proposed bonded joints under aging conditions.

The amphiphilic behavior of SDS [36,37] plays an important role in the aging properties of adhesive joints. For this reason, immersion tests have been carried out in CNT-doped adhesive films once cured by varying the amount of SDS. In addition to this, single lap shear (SLS) tests have also been conducted in standard coupons in order to see the effect of water and temperature aging. The main application of these reinforced joints is the SHM. The electrical response has been also monitored during these tests so that the electrical properties can be better characterized in order to obtain a deeper knowledge of the aging mechanisms.

#### **2. Materials and Methods**

#### *2.1. Materials*

The multi-wall CNTs used for this study were NC7000 supplied by Nanocyl, with an average diameter of 10 nm and a length of up to 2 μm.

The adhesive was a FM300K adhesive film, supplied by Cytec. This is an epoxy-based adhesive with a knit tricot carrier, which allows enhanced bondline thickness control. It has a high elongation and toughness, together with an ultimate shear strength of 36.8 MPa. It is suitable for bonding metal-to-metal and CFRP-to-CFRP systems.

CNT dispersion takes place by means of ultrasonication by using a previously optimized dispersion procedure [31]. It consists of a 20-min ultrasonication of a CNT aqueous solution at 0.1 wt%. The disaggregation of larger agglomerates is enhanced by the addition of a SDS surfactant. To study

the influence of this surfactant on the aging properties, the amount of SDS was fixed at 0.00, 0.25 wt% and 1.00 wt%.

After the dispersion procedure, the CNT suspension is sprayed over the adhesive surface prior to curing at a pressure of 1 bar at 40 cm for 0.5 s in order to achieve good homogenization of the CNTs over the surface.

In order to see the effects of aging, two types of specimens were prepared. One was the adhesive without substrate once cured—named in-bulk adhesive—in order to see the water uptake without any influence of the CFRP substrates. The second was the SLS specimens, which were made by secondary bonding of unidirectional CFRP substrates. The curing cycle was set for both the cured adhesive and the SLS joints in a hot press, as shown in Table 1. To improve the interfacial adhesion, the substrate surfaces were brushed.


**Table 1.** Cure-cycle parameters of secondary bonding.

#### *2.2. Aging Tests*

Cured adhesive and SLS specimens were subjected to aging conditions by immersion in distilled water at 60 ◦C, similar to some found in the literature [38]. Prior to immersion, the samples were dried in an oven at 50 ◦C for three days until weight loss was not observed between one and the next measurement. The aging time was set at 2 weeks (14 days) for the in-bulk adhesive, and up to 2 months (60 days) for SLS specimens. The reasons for these different immersion times were due to differences in the nature of each material and the exposed area subjected to water uptake. Adhesives tend to reach the water uptake saturation before the composite substrates [39] and in the case of the in-bulk adhesive, the exposed area of the adhesive is higher than in the CFRP joints. Therefore, the process of water uptake is accelerated [40].

Water absorption was measured in the in-bulk samples in their initial state and 1, 2, 3, 4, 7, 10 and 14 days after immersion. The water uptake was calculated by comparing the measured weight after immersion and the initial one in which is supposed that the samples were totally dry.

#### *2.3. Electromechanical Tests*

As commented before, bonded joints were subjected to SLS tests in order to study the aging effect on the electromechanical properties. They were conducted in three specimens for each condition (neat adhesive without CNTs and with 0.1 wt% CNTs with 0 wt%, 0.25 wt% and 1 wt% SDS). The tests were made according to standard ASTM D 5868-95 issue 01 using substrates of 100 × 25.4 × 2.5 mm with an overlapping area of 25.4 × 25.4 mm at a test rate of 13 mm/min. They were performed in a universal tensile Zwick machine.

Simultaneously, the electrical response was also monitored. Electrodes were made of copper wire sealed with silver ink in order to ensure a good electrical contact with the substrate surface. To protect the electrodes from environmental influences during testing, an adhesive layer was used. The measurements were carried out by an Agilent 34401A hardware and they were correlated to the mechanical response given by the tensile machine.

#### *2.4. Characterization*

Differential scanning calorimetry (DSC) measurements were conducted in a Metter Toledo mod 821 apparatus for the in-bulk adhesive. Two scans were carried out according to the standard ISO 11357-2:13, at 10 ◦C/min from ambient temperature to 250 ◦C. The glass transition temperature (*Tg*) was determined as the turning point of the heat capacity change. Two specimens of the non-aged- and

14-day-aged in-bulk specimens were measured in order to see the water uptake effect in the physical properties of the neat and CNT-doped adhesive.

#### **3. Results**

This section presents an analysis of the physical and mechanical evolution of SLS joints under aging conditions. First of all, the water uptake measurements for the in-bulk specimens are shown. Then, the mechanical properties of the SLS joints are discussed and finally, their electromechanical behavior is characterized.

#### *3.1. Water uptake Measurements*

Figure 1 shows the water uptake in terms of percentage of the initial weight for the in-bulk cured samples at each condition. The graph is in good agreement with the typical behavior of water uptake for this kind of samples, previously stated in other studies [38,41,42]. It is observed that water uptake is more prevalent in the initial stages and then the weight gain is going less significant until the water saturation is reached at 2 weeks of aging.

**Figure 1.** Water uptake graph for in-bulk specimens.

A similar water uptake behavior is found at every condition although some slight differences can be noticed. In this context, two opposite effects can play an important role. The first one is the hydrophobic behavior of carbon nanofillers, which can be introduced into the free volume of the polymer improving the barrier properties and leading to a reduction in the water uptake [43,44]. The second one is the amphiphilic behavior of the SDS that remains attached to the CNT surface [45,46], which can lead to an increase of the water absorption induced by the hydrophilic head groups [47]. In addition, CNT dispersion also plays a significant role. A poor dispersion can induce the presence of larger agglomerates, higher heterogeneity and higher distributed porosity. This irregular distribution leads, thus, to an irregular effect of the barrier properties of CNTs, which, in combination with the higher porosity, promotes a higher water uptake. However, a better dispersion of nanofillers improves the barrier properties leading, thus, to a lower water uptake.

The combination of these effects, as shown in the schematics of Figure 2, thus, explains the slight differences observed for each condition. In the case of the CNT reinforced adhesive without surfactant, a poor CNT dispersion is achieved, as stated in previous studies [31], so that the hydrophobic effect of CNTs is not so prevalent. Alternatively, the samples with 0.25 wt% and 1 wt% SDS show a similar trend, with a slightly lower water uptake than the sample without surfactant. In this case, the effect of the better CNT dispersion achieved was slightly prevalent over the amphiphilic effect of SDS surfactant. In the case of the neat adhesive, the water uptake was given directly by the physical behavior of the epoxy matrix.

**Figure 2.** Schematics of combined effect of the water uptake (**left**) and amphiphilic behavior of sodium dodecyl sulfate (SDS) (**right**).

#### *3.2. DSC Measurements*

Table 2 shows the *Tg* values at different testing conditions for the in-bulk specimens. In the initial state, a drastic reduction of *Tg* was observed when comparing the non-doped with the doped adhesive. This means that CNTs accelerated the curing process, leading to the maximum conversion point of the system (*Tg* ∼ 150 ◦C). This affirmation was supported by the measurements of the second scanning, where the *Tg* of all the samples were close to 150 ◦C, indicating the point of the maximum conversion of the resin. In addition, by observing the *Tg* of the aged samples, a significant decrease in comparison to the non-aged specimens was observed when adding CNTs, resulting in a similar glass transition temperature than for the neat adhesive. However, by observing the *Tg* obtained in a second scanning, it reaches the point of the maximum conversion in every case. This indicates that there was a plasticization effect caused by the water absorption in the case of CNT-doped samples. In the case of the neat adhesive, no significant differences were found when comparing aged and non-aged samples, so the plasticization effect was similar for the non-aged and aged samples. This affirmation was given by the fact that the network of the neat adhesive was not initially totally cured, with the plasticization effect less prevalent due to water absorption. It is important to note that the selected curing cycle was the same as that given by the supplier.


**Table 2.** Glass transition temperature for different in-bulk conditions.

Alternatively, when comparing the CNT-doped samples, it was observed that the addition of surfactant results in a more drastic reduction of the *Tg*, implying, thus, a higher plasticization effect. In order to better explain the possible effects that can take place in the material, it was necessary to focus on the mechanical testing of the SLS joints.

#### *3.3. Single Lap Shear Tests*

Figure 3 shows the lap shear strength (LSS) of the SLS specimens for each condition. It was observed that the LSS strength was significantly affected by the aging conditions. In every condition, a significant decrease of the LSS was observed after 1 month of aging while, for most of the cases, a slight recovery of the LSS was noticed by increasing the aging time. This different behavior can be

explained by attending the water uptake results and also by the different chemical interactions inside the adhesive joint.

**Figure 3.** LSS of adhesive joints for each testing condition.

First of all, it is necessary to understand the role of aging mechanisms due to water absorption. The water uptake generally induces the creation of weak hydrogen bonds, leading to a swelling of the polymer chains and causing, thus, a drastic reduction of the physical properties. This effect is called plasticization and is detrimental to the mechanical properties [41]. However, after this initial stage, a slight recovery is generally observed. This has already been shown in other studies [43,48,49] and can be explained by a change in the mechanism of water absorption. After reaching the water uptake saturation, longer aging times can induce a transformation of the weak hydrogen bonds into multiple chemical connections between the water molecules and the polymer chains, promoting an increase in secondary crosslinking, thus leading to a stiffening of the material and also an embrittlement, as observed in the examples given in Figure 4a. This cannot be confirmed by the *Tg* as there was only one measurement. Therefore, it is necessary to focus on the mechanical response of the adhesive joints, particularly on the displacement at failure.

**Figure 4.** Mechanical effect of water uptake showing (**a**) a plasticization stage and (**b**) the displacement at failure.

In the particular case of the neat adhesive, there was only a partial conversion of the epoxy matrix at the initial state, as the *Tg* was significantly below the maximum value (~150 ◦C). Therefore, the water uptake does not affect significantly the plastic properties of the epoxy matrix, as it presents a more prevalent plasticization effect at the initial state. This was confirmed by a slight variation of the displacement at failure with aging time, as shown in Figure 4b.

However, in the case of CNT-doped adhesive joints, there are some different mechanisms. Here, the CNT addition itself and the dispersion state, dominated by the surfactant addition, play

an important role in the aging mechanisms. As commented previously, a good dispersion implies a high homogeneity of CNT distribution in the matrix, improving the barrier properties. This acts in an opposite way to the amphiphilic behavior of SDS. In addition, the SDS also induces variations in the chemical interactions between the epoxy matrix and the CNTs, leading to a more drastic reduction of the *Tg*, as shown in Table 2. This was also observed by the displacement at failure shown in Figure 4b. In this case, the higher the SDS content, the higher the displacement at failure was. Moreover, it can be noted that the displacement at failure decreases after 2 months of aging. This was explained by the stiffening effect induced by the secondary crosslinking, which also leads to an embrittlement of the material, as previously stated. Therefore, the combination of the two effects explains the initial reduction of the LSS after 1 month and the general recovery for longer aging times.

In the case of the CNT sample without surfactant, a slight reduction of the LSS was observed after 1 month (~2%), while the effect on the mechanical properties was much more prevalent after 2 months, leading to a LSS reduction of more than 18%. In this case, the role of CNT dispersion was even more critical, as the absence of any surfactant leads to a much lower homogeneity of CNT distribution, inducing some areas with very high CNT content, acting as stress concentrators. Here, the embrittlement effect was more prevalent than the stiffening due to secondary crosslinking. This was confirmed by a higher relative reduction in the displacement at failure after 2 months of aging, similar to that initially obtained.

The higher displacement at failure observed in aging conditions for the samples with a higher amount of SDS can be explained accordingly to the presence of polar heads. They could interact with the matrix and the plasticizing effect is explained due to the unwinding of the chains around the nanotubes, as reported in other studies [50,51].

#### *3.4. Electrical Monitoring*

The aforementioned results give an initial idea of how aging conditions can affect the mechanical properties of CNT-doped adhesive film joints. In order to have a deeper understanding of aging effects on CNT-doped adhesive joints, it was necessary to analyze their electromechanical behavior. Figure 5 shows an example of electrical monitoring of a SLS specimen for different aging times. In every case, it was observed that the electrical resistance increases with displacement. This increase follows an approximately exponential behavior until failure, with the changes being more prevalent in the last stages of SLS testing. As stated in a previous study proving the monitoring capabilities of these CNT reinforced joints [28], the changes in the electrical resistance are due to the combination of two effects. The first effect was the increase of the tunneling distance between adjacent particles due to strain, leading to an increase of the tunneling resistance [52,53]. The second was the sudden crack propagation in the last stages of the tests, causing a prevalent breakage of electrical pathways through the joint. However, some important differences between the aged and non-aged specimens can be found.

By deeply analyzing the curves for the sample with 1 wt% SDS, the electrical behavior as a function of the applied strain changes was observed from the initial state to the 1-month-aged sample. In the aged sample, softer behavior was observed, due to the plasticization effect. This effect causes a steadier response of the material, with no abrupt changes in the electrical monitoring, as adhesive deformation and crack propagation take place in a softer way. By increasing the aging time, as discussed before, a secondary crosslinking was induced so the effect of plasticization was reduced, showing a mixed behavior between the initial state and the 1-month-aged sample, as noticed in the right graph of Figure 5c, where an abrupt change of the electrical behavior was observed.

The sample without surfactant shows similar behavior. At the initial state, some abrupt changes in the electrical resistance were observed, while the effect of plasticization was clearly shown after 1 month of aging (left graph of Figure 5b). By increasing the aging time, the stiffening effect of the secondary crosslinking was also observed by abrupt changes in the electrical behavior. In this case, as noticed before in the mechanical response, the behavior of the 2-month-aged sample was more similar to the non-aged one.

These initial results can give a good qualitative approximation of how aged and non-aged samples behave and prove the capability of CNT reinforced joints to properly monitor their mechanical behavior by means of electrical measurements. However, from the electrical response, it was possible to obtain estimations regarding damage evolution. To achieve this purpose, a simple analytical model, based on the tunneling effect of CNT reinforced polymers is proposed.

**Figure 5.** Electromechanical curves for single lap shear (SLS) specimens with (**left**) no surfactant and (**right**) 1 wt% SDS at (**a**) initial state and (**b**) after 1 and (**c**) 2 months of water immersion.

#### *3.5. Theoretical Approach*

The CNT-reinforced adhesive film was modeled as a nanocomposite with an homogeneous CNT distribution. The electrical mechanisms inside a nanoparticle network are given by the intrinsic electrical resistance of the CNTs themselves, the contact resistance between nanotubes and the tunneling resistance between near CNTs. In this particular case, the variation of electrical resistance due to applied strain are mainly dominated by the tunneling effect as intrinsic and contact resistance are assumed to be invariable with applied strain as stated in other studies [54].

Therefore, the changes in the electrical resistance can be divided into two terms, the first one, correlated to the variations due to the applied strain, which depends on the changes due to the tunneling effect between adjacent nanoparticles and the second one, which is correlated to the breakage of electrical pathways due to the effect of damage accumulation, as shown in the schematics of Figure 6, leading to the following expression:

$$
\Delta R\_{total} = \Delta R\_{tumnel} + \Delta R\_{dummy} \tag{1}
$$

**Figure 6.** Schematics of electromechanical behavior in a SLS test showing the increase of tunneling distance and crack nucleation inside the adhesive.

The tunneling effect can be calculated by using the well-known Simmons formula for the tunneling resistance [55]. It has been proved to be an effective way to predict the electromechanical response of a strained CNT network as it is the most dominant electrical transport mechanism. It takes several aspects such as the polymer type and the contact area between adjacent CNTs:

$$R\_{tunnel} = At \cdot e^{bt} \tag{2}$$

where *A* and *b* are two constants depending on the CNT geometry and matrix barrier characteristics and *t* is the tunneling distance, which changes with the applied strain.

The changes due to damage accumulation are not easy to model. There are many studies investigating this effect by proposing different damage evolution laws [56,57], but the particularities of the tested systems make the damage calculation very difficult. Therefore, damage accumulation is estimated by comparing the measured changes in the electrical resistance and the known tunneling effect.

$$
\Delta R\_{damage} = \Delta R\_{measured} - \Delta R\_{tunvel} \tag{3}
$$

For this purpose, the initial tunneling distance is calculated as the distance that best fits the initial changes of the electrical resistance, where no damage is supposed.

Figure 7 shows the comparison between the theoretical line, taking only the tunneling effect and the experimental measurements for aged and non-aged samples with 1 wt% SDS into account. The pattern areas indicate the differences between the theoretical and the experimental ones being, thus, the damage accumulation during the SLS test.

**Figure 7.** Damage accumulation evolution (red pattern area) by comparing experimental measurements (solid lines) and theoretical predictions (dashed lines) for 1 wt% SDS samples at (**a**) initial state and (**b**) after 1 and (**c**) 2 months of aging.

The non-aged sample shows significant irregular behavior. The threshold for damage accumulation was observed at ~0.25 mm displacement. After this point, the evolution of damage accumulation was very irregular. This can be explained due to the brittle mechanisms dominating the mechanical behavior of the adhesive. Secondary cracks start to nucleate and then they coalesce around the main crack [58], inducing a higher breakage of electrical paths in a similar way than that observed in other studies for fatigue testing [59]. This nucleation was not uniform so the unstable damage accumulation was explained. In the case of 1-month-aged samples, this damage accumulation starts to take place at 1 mm of displacement, that is, much later than in the non-aged specimen. This is in good agreement with the stated conclusions from water uptake and LSS measurements, as the induced plasticity in the first stages of water uptake avoids the early crack nucleation inside the adhesive joint. After that, damage accumulation takes places in a sudden way, that is, the cracks start to nucleate and then immediately coalesce. The 2-month-aged sample has a damage threshold of 0.5 mm, lower than in the 1-month-aged sample, due to the stiffening effect of the change in the water absorption mechanisms discussed above. Then, a softer evolution of damage accumulation is observed and finally, in the last stages of SLS tests a rapid coalescence takes place until final failure.

The previously described behavior was similar in the case of the 0.25 wt% SDS samples. However, in the case of the samples without surfactant, the electromechanical behavior shows some slight differences regarding the 1 wt% SDS samples, especially, concerning the sensitivity of the electrical response. Figure 8 presents the comparison between the theoretical and the experimental lines at different aging times. At the initial state, similar behavior for the 1 wt% SDS samples was observed with abrupt changes in the electrical resistance, inducing a high damage accumulation rate due to the

rapid nucleation and coalescence of micro-cracks inside the material. The 1-month-aged specimen shows a much softer behavior, as expected due to the plasticization effect of the water uptake process. The threshold of damage accumulation was observed at a 0.7 mm displacement, that is, much later than for non-aged samples. However, unexpected behavior was observed for the 2 month-aged sample. In this case, the threshold for damage accumulation was observed nearly at the beginning of the SLS tests, that is, earlier than in the case of non-aged specimens. In addition, the damage accumulation was very high also in comparison to the other specimens as they show a much higher sensitivity.

**Figure 8.** Damage accumulation evolution (red pattern area) by comparing experimental measurements (solid lines) and theoretical predictions (dashed lines) for non-surfactant samples at (**a**) initial state and (**b**) after 1 and (**c**) 2 months of aging.

This behavior can be explained by the interaction of two effects. One of them was due to the fact that the highly heterogeneous CNT distribution can induce high stress concentrations, leading to highly weak points in the matrix. This effect was also present in the non-aged and aged specimens, but in the case of the 2-month-aged one, the stiffening effect induced by the change in the mechanism of water absorption previously described can lead also to much higher embrittlement of the adhesive. This can also be stated in the reduction of the displacement at failure, so that the nucleation and coalescence of microcavities takes place more rapidly. The second one was correlated to the different interactions between the larger agglomerates of CNTs (which were much more prevalent in the non-surfactant samples) and the water molecules. This promotes a different electromechanical behavior than for the other samples.

#### **4. Conclusions**

SLS joints with CNT-doped adhesive films have been tested under aging conditions while their electromechanical properties have been monitored.

The water uptake measurements for the cured adhesive without substrate show that the behavior does not change significantly with the addition of both CNTs and SDS surfactant. This is explained by the combined effect of the amphiphilic behavior of the SDS and the barrier properties of CNT dispersion, acting in an opposite way.

The LSS of the bonded joints shows a general decrease after 1 month of aging because of the plasticizer effect of the water, which promotes the creation of weak hydrogen bonds. This statement was also confirmed by an increase of the displacement at failure. After 2 months of aging, there was a slight increase of LSS and a general reduction on the displacement at failure, which was explained by the secondary crosslinking that takes place due to water uptake saturation. In the case of the sample without surfactant, this behavior was slightly different because of the poor CNT distribution that can induce higher embrittlement, leading to a sudden decrease of LSS even after 2 months of aging.

Finally, the analysis of the electromechanical behavior of SLS joints confirms the previously described statements. A higher plasticization was observed for 1-month-aged specimens, while a partial recovery of the stiffness was observed after 2 months. By comparing the measured electrical response with simple theoretical calculations, it is possible to obtain the first quantitative idea of damage accumulation and how aging conditions affect the damage evolution. Therefore, this first estimation can be used to better understand the physical mechanisms taking place on CNT-doped adhesive joints under aging conditions subjected to SLS tests.

As a future work, however, it would be necessary to refine the theoretical predictions by taking some effects such as CNT distribution and orientation or the barrier properties of the epoxy matrix into account. This could give more accurate knowledge of the physical behavior of CNT-doped bonded joints under aging conditions.

**Author Contributions:** X.F.S.-R. conceptualization, methodology, formal analysis, writing—original draft preparation; A.J.-S. conceptualization, writing—review; S.G.P. supervision, formal analysis, writing—review; M.S. conceptualization; A.G. and A.U. funding acquisition. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Ministerio de Economía y Competitividad of Spanish Government (Project MAT2016-78825-C2-1-R) and Comunidad de Madrid regional government [PROJECT ADITIMAT-CM (S2018/NMT-4411)].

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2020 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/).

### *Article* **Decomposition of Flavonols in the Presence of Saliva**

#### **Malgorzata Rogozinska and Magdalena Biesaga \***

Faculty of Chemistry, University of Warsaw, Pasteura 1, 02-093 Warsaw, Poland; mgwiazdon@chem.uw.edu.pl **\*** Correspondence: mbiesaga@chem.uw.edu.pl

Received: 23 September 2020; Accepted: 22 October 2020; Published: 26 October 2020

**Abstract:** In this study, the LC-MS/MS was applied to explore the stability of four common dietary flavonols, kaempferol, quercetin, isorhamnetin, and myricetin, in the presence of hydrogen peroxide and saliva. In addition, the influence of saliva on the representative quercetin glycosides, rutin, quercitrin, hyperoside, and spiraeoside was examined. Our study showed that, regardless of the oxidative agent used, flavonols stability decreases with increasing B-ring substitution. The decomposition of analyzed compounds was based on their splitting by the opening the heterocyclic C-ring and realizing more simple aromatic compounds. The dead-end products corresponded to different benzoic acid derivatives derived from B-ring. Kaempferol, quercetin, isorhamnetin, and myricetin were transformed into 4-hydroxybeznoic acid, protocatechuic acid, vanillic acid, and gallic acid, respectively. Additionally, for quercetin and myricetin, two intermediate depsides and 2,4,6-trihydroxybenzoic acid derived from A-ring were detected. All analyzed glycosides were resistant to hydrolysis in the presence of saliva. Based on our data, saliva was proven to be a next oxidative agent which leads to the formation of corresponding phenolic acids. Hence, studies on flavonols' metabolism should take into consideration that the flavonols decomposition starts in the oral cavity; hence, in subsequent parts of the human digestive tract, they could be present not in their parent form but as phenolic acids. Further analyses of the influence of saliva on flavonols glycosides need to be performed due to the possible interindividual fluctuations.

**Keywords:** flavonols; H2O2; saliva; metabolism; oxidation; LC-MS/MS

#### **1. Introduction**

Flavonoids are phenolic compounds widely present in plants and food of plant origin. Both clinical and epidemiological studies show the correlation between the dietary polyphenols intake and the reduction of risk of some chronic diseases such as cardiovascular diseases, cancer, and diabetes, as well as aging [1–4]. These beneficial effects are associated with their antioxidant activity [5–7]. Flavonols, the major class of flavonoids present in the human diet, and among them quercetin, kaempferol, myricetin, and isorhamnetin (Figure 1), are well known to act as antioxidants in vitro and show protective effects against free radicals, reactive oxygen species, and other oxidation agents [8–10]. However, the biological properties of antioxidants such as flavonols depend on their bioavailability and metabolism in the human body. The study of biological responses due to dietary intake of polyphenols cannot be carried out without taking into consideration polyphenols–saliva interactions. Although phenolics are normally ingested through the mouth as elements of food, very little is known about their metabolism in the oral cavity. One of the main goals assigned to saliva is participation in glycosides hydrolysis [11,12], which delivers biologically active aglycones that can be absorbed more effectively in the human digestive tract. Since flavonols mainly occur in food as glycosides, much of the research focused on the metabolism in oral cavity concerns glycosides. However, some reports show the presence of flavonols aglycones in the food of plant origin, for instance in eucalyptus and unifloral types of honey [13]. Moreover, the fermentation carried out during the manufacturing of food could results in the hydrolysis of glycosides to aglycones. Although it was reported that flavonoid glycosides

can be absorbed intact via the sodium-dependent glucose transporter [14], it was also shown that many glycosides are not absorbed due to efficient efflux transport by intestinal efflux protein pumps [15].

**Figure 1.** Structures of kaempferol, quercetin, isorhamnetin, and myricetin.

It is well established that the oral cavity harbors numerous and diverse microorganisms, which can hydrolyze flavonols glucosides to the aglycones by glucosidases excreted from the bacteria [11,16]. All of these features indicate that the effective absorption of polyphenols in the human digestive tract strongly depends on their deglycosylation, and, as far as we know, study on the metabolism of aglycones in the oral cavity are limited. Interaction of saliva components with bioactive compounds from food occurs due to various reasons. Firstly, chewing food allows its components to stay in the oral cavity for a while, which ensures the contact of saliva and oral mucosa with food. Secondly, during chewing, flavonols are dissolved in saliva, which facilitates their interaction with oral mucosa and salivary proteins [17–20]. It is worth noting that Quijada-Morín et al. [17] outlined interactions between flavan-3-ols and salivary proteins not only as a precipitation issue as it has been usually studied but also as a more complex interaction, which involves the formation of soluble and insoluble complexes.

All of these mechanisms increase lipophilic polyphenols assimilation and causes their retention in the oral cavity over time [21–23].

The human oral cavity contains numerous and diverse microorganisms as commensals [24,25]. Approximately 280 bacterial species from the oral cavity have been isolated in culture and formally named. The possible reason for the decomposition of flavonols in the presence of saliva is the fact that bacteria and leucocytes presented in the oral cavity are able to generate H2O2; thus, flavonols could be easily oxidized [26,27]. While there is a lack of reports showing the oxidation of flavonols in the presence of saliva, other oxidizing conditions have been well established, mainly for quercetin. Quercetin is degraded in air conditions with the formation of different depsides, as intermediates in the degradation pathway. Further decomposition results in the formation of 2-(3,4-dihydroxyphenyl)-2-oxoacetic acid, 2,4,6-trihydroxybenzoic acid, and 3,4-dihydroxybenzic acid [28]. Additionally, Maini et al. [29] proposed another degradation products of quercetin after its exposure to UVA radiation: 1,3,5-trihydroxybenzene and 2,4,6-trihydroxybenzaldehyd. In general, oxidation of quercetin by various methods: air, electrochemical, enzymatic, and free radical oxidation may yield, more or less, the same set of oxidized products [30].

For further elucidation of the oxidation processes of flavonols in biological systems, we investigated the stability of four wide-spread flavonols: quercetin, kaempferol, myricetin, and isorhamnetin in the presence of saliva. This work also provides comparative oxidative studies of flavonols using H2O2 solution and whole saliva as oxidation agents. We chose flavonols that differ in the number of hydroxyl substituents in the B-ring. Additionally, in the case of isorhamnetin, one hydroxyl group is replaced with a methoxy group, which allows for the blocking of interactions from two adjacent hydroxyl groups. To determine the various degradation products of flavonols, LC-MS/MS system was applied.

#### **2. Materials and Methods**

#### *2.1. Chemicals and Reagents*

The commercial standards of flavonols and phenolic acids, as well as the rest of the chemicals, were purchased from Sigma-Aldrich (Steinheim, Germany). Glycosides rutin (quercetin 3-rhamnoglucoside), quercitrin (quercetin 3-rhamnoside), hyperoside (quercetin 3-galactoside), and spiraeoside (quercetin 4 -glucoside) were purchased from Extrasynthese (Genay, France). Methanol was obtained from Merck (Darmstadt, Germany). In all experiments, ultrapure water from a Milli-Q system with an electrical resistivity of 18 MΩ × cm was used. Stock solutions of flavonoids were prepared in methanol. Diluted standards were prepared in 25% methanol with a final concentration of 50 mg/L. All solutions were filtered through 0.45 μm membranes (Millipore) and degassed prior to use.

#### *2.2. Collection of Saliva Samples*

Five healthy human volunteers (20–30 years old) were recruited to this study based on restricted criteria. All volunteers were self-reported to be in good general health without any chronic diseases, not taking antibiotics, and no history of drinking and smoking habits. Volunteers were requested to refrain from eating for at least 10 h and they could drink only mineral water before collection. Saliva samples were collected in the morning 2 h after brushing teeth with toothpaste free of polyphenols. Before saliva collection, the oral cavity was rinsed with ultrapure water to remove dead skin cells. Saliva was mechanically stimulated by chewing a plastic tube and collected into the plastic containers. Then, individual saliva samples (n = 5) were diluted with water (1:1, w:w) and incubated at 37 ◦C in a water bath until analysis.

#### *2.3. Influence of Saliva*/*H2O2 Solution on Stability of Flavonols*

Standards of kaempferol, isorhamnetin, quercetin, and myricetin were mixed with prepared saliva or H2O2 solution at 25 mg/L in 12.5% of MeOH final concentration and incubated for 24 h at 37 ◦C. Samples were prepared to obtain the ratio 1:1 v:v of saliva or equimolar H2O2 to the standard solution. The ratio of saliva and standard solution was evaluated to obtain optimal flavonols decomposition time, which allowed detecting unstable intermediates [31]. Since the non-glycosylated flavonols did not easily dissolve in the water, addition of MeOH had to be used, and the minimum concentration of MeOH was 25% in stock solution (12.5% in sample solution). Additionally, each flavonol was mixed with the same volume of the water, as a control sample. All samples were filtered through 0.45-μm PTFE membranes and analyzed at different times of incubation over 24 h. A long incubation period allowed us to find the dead-end products of flavonols degradation. The pH value of all solutions was controlled and measured at the beginning and end of the reaction.

#### *2.4. Influence of Microbiota*

To check if the decomposition of flavonols is the result of the presence of microbiota, we decided to incubate quercetin (as a representative flavonol) with saliva filtered using sterile polyethersulfone (PES) membrane (pore size 0.22 μm) to remove bacteria from the sample. For that purpose, 5 mL of saliva was diluted 1:1 with distilled water and shaken vigorously to reduce viscosity. In parallel, 100 μL of filtered saliva was spread onto 5% blood agar and incubated for 24 h at 37 ◦C in duplicate to examine if the usage of 0.22 μm membranes ensures the sterility of the sample. At the same time, the experiment with the incubation of quercetin with bot saliva filtered using 0.22 and 0.45 μm membranes was carried out.

#### *2.5. Apparatus*

The analytical method used in the presented work was developed in our laboratory and discussed in detail in a previous paper on the study of the phenolic compounds [32]. Chromatographic analyses were carried out using LC-MS/MS system consisted of binary pumps LC20-AD, degasser DGU-20A5, column oven CTO-20AC, and autosampler SIL-20AC, connected to 3200 QTRAP Mass spectrometer (Applied Biosystem/MDS SCIEX). Compounds were separated on Chromolith Performance C18 column (100 × 2mm, Merck) at 30 ◦C. Formic acid (8 mM, pH 2.8) as eluent A and methanol as eluent B were used. Samples incubated at 37 ◦C were kept in the same temperature before analysis in an autosampler. The flow rate of mobile phase was 0.2 mL/min and the gradient mode was as follow: 0–3 min, 10% B; 20–25 min, 50% B; 26–40 min, 10% B. LC system was connected to the 3200 QTRAP Mass spectrometer (Applied Biosystem/MDS SCIEX) with electrospray ionization (ESI) working in negative mode. ESI conditions were as follows: capillary temperature of 450 ◦C, curtain gas at 0.3 MPa, auxiliary gas at 0.3 MPa, and negative ionization mode source voltage of 4.5 kV. Nitrogen was used as curtain and auxiliary gas. Analyst 1.4.2 software was used for data acquisition. LC-MS/MS analysis were carried out by comparing retention time and *m*/*z* values obtained by MS and MS2 with the mass spectra of standards obtained under the same conditions. Because some degradation products such as depsides are not commercially available, the presence of these compounds was confirmed by comparison of retention times, masses, and fragmentation spectra of potential oxidation products with literature.

Quantification of compounds was done using the selected reaction monitoring mode (SRM). For each compound, the optimum conditions for SRM mode were determined in infusion mode and two SRM pairs were chosen as representatives (SRM1 and SRM2) (Table 1). Due to the higher intensity of peak obtained using the SRM1 pairs, they were chosen for qualitative analyses. Calibration curves were drawn from the analysis of 5 μL volumes at concentration ranging from 0.5 to 50 mg/L (n = 7) measured in triplicated. Coefficients of linearity (R2) for the calibration curves were <sup>≥</sup> 0.996. LODs were estimated by decreasing the concentration of the analytes to the smallest detectable peaks, and then its concentration was multiplied by three. LODs ranged 0.1–0.5 mg/L.


**Table 1.** LC-MS/MS characteristics of phenolic compounds in the negative mode.

#### *2.6. Statistical Analysis*

The statistical analyses of the data were carried out using Microsoft Excel 2016 and Excel's Analysis Toolpak (ANOVA). The one-way analysis of variance (ANOVA) and the significance of differences between sample means were calculated, and *p* values ≤ 0.05 were taken into account as significant.

#### **3. Results and Discussion**

#### *3.1. Oxidation of Flavonols with Hydrogen Peroxide*

As compared to the control, a noticeable decrease of quercetin concertation was observed during its incubation in the presence of H2O2. As shown in Figure 2, rapid degradation of quercetin could be observed with the simultaneous formation of three other compounds with [M-H]− at *m*/*z*317 (tr = 17.5 min) [M-H]<sup>−</sup> at *m*/*z* =193 (tr = 4.4 min) and [M-H]<sup>−</sup> at *m*/*z* 169 (tr = 5.0 min). Oxidation product with [M-H]− at *m*/*z* 317, identified as depside characteristic for quercetin, was found as an intermediate product, the concentration of which increased in the first 80 min of the experiment. Further incubation led to the decrease of its concentration, whereas new peaks with [M-H]<sup>−</sup> at *m*/*z* 193 (tr = 4.4) and [M-H]<sup>−</sup> at *m*/*z* 169 (tr = 5.0) appeared. Final degradation products were identified as 2,4,6-trihydroxybenzoic acid (*m*/*z* 169) and protocatechuic acid (*m*/*z* 193). Such results suggest that the oxidation of quercetin with H2O2 is based on its splitting by the opening the heterocyclic C-ring and realizing simpler aromatic compounds. Moreover, oxidation involves the initial oxidative step with subsequent changes in the flavonol skeleton such as the formation of B-ring orthoquinone and rearrangement in the C-ring [33]. Detected oxidation products are similar to those obtained under other conditions such as oxygen [28], UVA and UVB [34], hydroxyl free radical [35], and presence of copper (II) [36,37]. Thus, the hypothesis of Zhou et al. [30] that the oxidation of quercetin using different oxidations agents may yield, more or less, the same set of oxidized products seems to be particularly relevant. Decomposition pattern of myricetin leads throughout the characteristic depside as an intermediate at *m*/*z* 321. The LC-MS/MS measurements show that further oxidation led to its decomposition and formation of gallic acid as a corresponding hydroxybenzoic acid derivative at *m*/*z* 169 and tr = 2.9 min. Besides, 2,4,6-trihydroxybenzoic acid (tr = 4.9 min) was detected as a degradation product of myricetin. Although it shares the same mass as gallic acid, their separation was obtained in the established method. Unfortunately, neither for isorhamnetin nor for kaempferol corresponding depsides were detected. Nevertheless, transformations which involve initial oxidative steps with subsequent changes in the flavonols skeleton was observed. As a result of kaempferol oxidation, the formation of a compound with tr = 7.1 min and SRM characteristic for 4-hydroxybenzoic acid (137/93) was observed. The breakdown product of isorhamnetin was identified as a vanillic acid with retention time 9.8 min and SRM pair (167/152). These results suggest that, as in the case of quercetin and myricetin, oxidation of kaempferol and isorhamnetin leads to C-ring cleavage in the flavonols' structure. As a result of this reaction, the corresponding hydroxybenzoic acid derivatives are formed. It should be mentioned that two reactive centers in the C-ring were identified as responsible for antioxidant activity in flavonols: the 2,3 double bond in conjugation with the 4-oxo function and the 3- and 5-hydroxyl groups with hydrogen bonding to the same 4-oxo function (for flavonols numbering, see Figure 1) [38,39]. That is why all of the modifications which occur in that area could significantly alter the chemical and biological properties of flavonols.

**Figure 2.** Degradation profile of quercetin and its main products formed during the incubation with H2O2 solution.

#### *3.2. Stability of Flavonols in Saliva Solution*

Figure 3 presents the loss of the starting amount as percent of remaining kaempferol, quercetin, and isorhamnetin (Figure 3a) as well as myricetin (Figure 3b) for a single representative after 6 h of incubation with saliva. The levels of the examined compounds were significantly different after 6 h of incubation. Kaempferol was the most stable compound in saliva during this time (about 73% left). Quercetin and isorhamnetin were less stable than kaempferol, and the degree of their degradation was 52% and 59%, respectively. Myricetin was the least stable flavonol under these conditions, and its concentration was at trace levels at the end of the experiment. As can be readily seen, myricetin showed rapid degradation and disappeared after 90 min of incubation. The general order of stability of examined flavonols was as follows for all saliva samples collected from five human volunteers: kaempferol > isorhamnetin > quercetin >> myricetin. Generally, the decomposition rate of flavonols increases with an increasing number of hydroxyl groups attached to the B-ring. A similar order of degradation of these flavonols was observed in H2O2 solution. These results are in good agreement with observations of Maini et al. for ultraviolet radiation A (UVA) [29]. During the incubation of quercetin with saliva solution, the decomposition of quercetin into compounds with *m*/*z* 305 and *m*/*z* = 317 was observed. These two peaks were identified as the intermediate products of quercetin oxidation (depsides) and their formation under oxidative conditions has already been described [28,29,33]. During further incubation, depsides decomposed with a simultaneous appearance of another peak with the retention time 4.41 min, mass spectrum, and SRM pair (153/109) characteristic for protocatechuic acid (Figure 4). In contrast to the incubation of quercetin in the presence of H2O2 mixture, the formation of 2,4,6-trihydroxybenzoic acid as a degradation product was not observed. Similar results were obtained for myricetin. However, the latter was less stable in the presence of saliva than quercetin. Chromatogram of myricetin incubated with saliva showed a decrease in its peak's intensity, whereas a new peak at *m*/*z* = 321 appeared. This peak was considered as depside characteristic for myricetin. Moreover, the formation of a new compound with retention time (tr = 2.9 min), mass spectrum, and SRM pair characteristic for gallic acid (169/125) was observed. Unfortunately, as mentioned above, among studied flavonols, only for myricetin and quercetin the corresponding depsides were detected. Chromatograms and mass spectrum obtained for isorhamnetin and kaempferol showed their degradation to vanillic and 4-hydroxybenzoic acids, respectively. In an oxidation experiment, Maini et al. [29] suggested that the presence of the corresponding phenolic acid derivative in the absence of any detectable depside concentration is the result of comparable depside formation and hydrolysis rates. Krishnamachari et al. suggested that the presence of both a catechol unit in the B-ring and a free C-3 hydroxyl appears to be a prerequisite for the formation C-ring carbocation or *p*-quinone methide (which formation proceeds predominantly through its tautomer, *o*-quinone) in the oxidative decomposition of flavonols [33,40]. It has been also demonstrated using EPR spectroscopy that the spin distribution during oxidation of quercetin remains entirely on the B-ring, promoting the donation of two electrons leading to the formation of an *o*-quinone [41]. These phenomena explain our observations that myricetin and quercetin decompose rapidly and respective depsides could be observed. Moreover, hydroxyl or methoxy substituents are considered to stabilize the flavonol C-ring carbocation intermediate. As has already been proven, a relatively electron-rich derivative may be more stable and hence more easily formed than its electron-poor analog [42]. Hence, electron-withdrawing and electron-donating groups such as hydroxyl and methoxy groups attached directly to B-ring should influence the rate of *p*-quinone methide formation. In addition, two or more electron-donating groups greatly facilitate its initial generation and stability [42]. Additionally, isorhamnetin due to the presence of a methoxy group greatly enhances the electron-donating properties in the 4'-position [38]. Finally, the presence of corresponding benzoic acid derivatives, as well as 2,4,6-trihydoxybenzoic acid, highlights possible interconversion of the *p*-quinone methide into C-ring carbocation intermediate and its further decomposition. Poorly substituted flavonols such as kaempferol with one hydroxyl group in B-ring skeleton are not able to form *o*-quinone or its tautomeric form *p*-quinone methide species readily. This theory could also explain why any detectable levels of depsides were observed for kaempferol and isorhamnetin.

**Figure 3.** The loss of the starting amount in saliva solution as percent of remaining: (**A**) kaempferol, quercetin, and isorhamnetin; and (**B**) myricetin.

**Figure 4.** Changes in amount of quercetin (22.8 min) and protocatechuic acid (4.5 min) during incubation of quercetin with saliva.

As mentioned above, saliva was proven to contain phenolics after consumption of phenolics-rich beverages, and these compounds have been found to persist in oral cavity up to 300 min, despite a constant salivary flow [21]. Our research shows that this time is sufficient for the partial decomposition of kaempferol, quercetin, and isorhamnetin and complete decomposition of myricetin. To check if detected compounds may be considered as a dead-end product, we decided to incubate analyzed flavonols for further 18 h. After 24 h of incubation, no additional compounds were detected as degradation products of the analyzed flavonols. However, further decomposition of kaempferol, quercetin, and isorhamnetin was observed. The levels of remaining flavonols calculated as average amount for five volunteers was as follow: 55.42% ± 8.12 of kaempferol, 23.17% ± 5.48 of quercetin, and 39.37% ± 11.11 of isorhamnetin. This indicates that, even after long time of incubation, the order of stability of flavonols remains the same. Since it is known that the stability of flavonoids strongly depends on the pH, we controlled it during the experiment. For all solutions, pH fluctuation ranged from 6.96 ± 0.10 to 7.06 ± 0.15 at the beginning and from 6.98 ± 0.12 to 7.39 ± 0.09 at the end of the experimental period.

Several studies have reported that saliva can hydrolyze flavonoids glycosides, hence we checked four quercetin glycosides: rutin (quercetin 3-rhamnoglucoside), quercitrin (quercetin 3-rhamnoside), hyperoside (quercetin 3-galactoside), and spiraeoside (quercetin 4 -glucoside). Preliminary studies showed that all of the analyzed glycosides were stable in the presence of saliva. The results obtained for rutin and quercitrin are in good agreement with previous studies, which showed that these two glycosides were hydrolyzed very slowly or were resistant to salivary hydrolysis [12,31]. On the other hand, lack of spiraeoside hydrolysis was inconsistent with experiments which suggested that glucose conjugates are rapidly hydrolyzed to corresponding aglycones [11,12]. However, in the same study, a large interindividual variability in hydrolysis rate was also observed. According to the aim of our study, it is interesting to note that sugar moiety attached to flavonol inhibits C-ring decomposition.

The presented studies allow creating the general scheme of flavonols' degradation in the presence of two oxidation agents: H2O2 and saliva (Figure 5). Table 2 presents the summarized products detected in samples after oxidation with H2O2 and saliva.

**Figure 5.** General scheme of flavonols' degradation in the presence of saliva.



− not detected; + detected.

Overall, decomposition of flavonols is based on the cleavage of heterocyclic C-ring, with no changes in hydroxyl and methoxy substituents in A-ring and B-ring. Decomposition of myricetin and quercetin leads to the formation of gallic and protocatechuic acid, respectively, which exhibit high redox potential due to the presence of adjacent hydroxyl groups attached to the aromatic ring [43]. Contrarily, vanillic and 4-hydroxybenzoic acids were found to be less efficient in radical neutralization reaction [43].

As mentioned above, the human oral cavity contains numerous and diverse microorganisms as commensals. It is known that microbiota can split the flavonoids by opening the heterocyclic ring and releasing simpler aromatic compounds, such as hydroxyphenylacetic acids from flavonols, which could be further metabolized to derivatives of benzoic acid [44]. Taking that into consideration, we decided to incubate quercetin, as a representative flavonol, with saliva containing microorganisms (filtration through 0.45 μm filters) and without microorganisms (filtration through 0.22 μm sterile filters). In both cases, quercetin was degraded to the same extent during its incubation to protocatechuic acid, as a dead-end product. This indicates that the decomposition of quercetin is independent of the presence of microbiota.

#### **4. Conclusions**

As noted in this study, the stability of flavonols in the presence of saliva solution strongly depends on the number of hydroxyl groups attached to the B-ring. Indeed, flavonol stability decreases with increasing B-ring substitution. This proves that saliva is a next oxidative agent, besides UVA and UVB radiation, air, enzymes, and free radicals, which leads to the formation of corresponding phenolic acids as dead-end products of flavonols. The obtained results indicate that myricetin is the most effective flavonol in the process of neutralizing free radicals formed in the oral cavity, due to its easy oxidation caused by the presence of three hydroxyl groups in the B-ring of this compound. Quercetin and isorhamnetin can also be regarded as quite effective antioxidants, capable of oxidizing in the presence of saliva. Nevertheless, their degradation products such as gallic acid and protocatechuic acid still possess reducing potential and are well-known antioxidant units. Rutin (quercetin 3-rhamnoglucoside) and quercitrin (quercetin 3-rhamnoside), as well as, surprisingly, hyperoside (quercetin 3-galactoside) and spiraeoside (quercetin 4 -glucoside), were resistant to hydrolysis in the presence of saliva. Nevertheless, due to the interindividual fluctuations, further analyses should be elucidated on that issue. However, the most intriguing seems to be the relatively short residence time of most foods and their bioactive compounds in the oral cavity. Even so, it should be noticed that this time is sufficient for flavonols decomposition, which clearly shows the process of flavonols metabolic transformation starts in the oral cavity. Hence, studies on flavonols metabolism should take under consideration that, in subsequent parts of the human digestive tract, they could be present not in their parent form but as phenolic acids. Consequently, information on the bioavailability and metabolic pathway of such dietary bioactive compounds is a key part in understanding their beneficial influence on human health.

**Author Contributions:** Conceptualization, M.R. and M.B.; methodology, M.R., M.B.; validation, M.R., M.B. investigation, M.R., M.B.; writing—original draft preparation, M.R.; writing—review and editing, M.B.; supervision—M.B.; All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Acknowledgments:** The presented studies were carried out with the approval of Ethics and Bioethics Committee (KEIB–5/2016, Cardinal Stefan Wyszynski University in Warsaw). LC-MS/MS measurements were performed at the Laboratory of Structural Research, Faculty of Chemistry, University of Warsaw, which was established under the European Regional Development Grant WPK\_1/1.4.3./2004/72/72/165/2005/U. The authors thank prof. Dorota Korsak for her help in the microbiological part of the presented manuscript and prof. Krystyna Pyrzy ´nska for her helpful advice on various technical issues examined in this paper.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 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/).

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Applied Sciences* Editorial Office E-mail: applsci@mdpi.com www.mdpi.com/journal/applsci

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18