VOC Profiles of Saliva in Assessment of Halitosis and Submandibular Abscesses Using HS-SPME-GC/MS Technique

Halitosis and submandibular abscesses are examples of mouth-related diseases with the possible bacterial origin. Salivary volatile organic compounds (VOCs) are potential biomarkers of them, once they can be addressed as metabolites of bacterial activity. Healthy patients (n = 15), subjects with submandibular abscesses located in fascial deep space (n = 10), and subjects with halitosis (n = 5) were enrolled in the study. Saliva samples were subjected to headspace solid-phase microextraction (HS-SPME) and gas chromatography coupled to mass spectrometry (GC/MS) analysis. A total number of 164 VOCs was detected by the developed methodology, 23 specific for halitosis and 41 for abscess. Halitosis’ profiles were characterized by a larger number of sulfur compounds, while for abscess they had a higher variety of alcohols, aldehydes, and hydrocarbons—biomarkers of inflammatory processes. Principal components analysis allowed visualization of clusters formed according to the evaluated conditions. Kruskal-Wallis test indicated that 39 VOCs presented differentiated responses between the studied groups, with statistical relevance (p < 0.05). Random forest was applied, and a prediction model based on eight VOCs (2-butanone, methyl thioacetate, 2-methylbutanoic acid, S-methyl pentanethioate, dimethyl tetrasulfide, indolizine, pentadecane, and octadecanal) provided 100% of sensitivity, 82% of specificity, and 91% of balanced accuracy, indicating the specific presence of submandibular abscess.


Extraction time
Different extraction times (10,30,45 and 60 min) were tested in triplicate for a pool of saliva, prepared as described in 1.1.. Figure 2 presents obtained average total area (summed up areas of detected peaks) for each time tested.

A graph depicted in
Extraction conducted during 45 min displayed acceptable performance and lower standard deviation was observed, indicating better reproducibility of profiles.

Sample volume
Different volumes of sample (0.5, 1, 2 and 3 mL) were tested in triplicate for a pool of saliva, prepared as described in 1.1.. Figure 3 presents obtained average total area (summed up areas of detected peaks) for each tested volume, as well as the average number of detected peaks. 0.5 mL and 1 mL presented to be the most suitable volumes for saliva sample analysis, since larger number of peaks could be detected. Since no analytical disadvantage was verified in the present case, the use of lower amount of biological specimen was prioritized.

Fresh saliva vs. incubated saliva
Aliquots of fresh saliva samples were compared to samples submitted to incubation at 37°C, for 24 h. These samples correspond to those enrolled in this study (n = 30). The results were expressed in terms of average total area (average of sum of peak areas obtained for a group of samples) and number of detected peaks. The Figure 4 depicts these observations, highlighting the obtaining of richer profiles when incubation step is incorporated.

Figure 4-Comparison between fresh and incubated saliva, in terms of (A) average total area and (B) number of detected
peaks.

Internal validation
In order to verify the suitability of the developed method to its purpose, a simplified validation process was carried out. The following parameters were evaluated: precision, accuracy, linearity, carry over, matrix effect and stability. Once the present work is based in non-target metabolomics and endogenous compounds are being investigated, 26 VOCs standards were used as model analytes and spiked to blank samples (0.5 mL of deionized watersimulating the major content of the biological sample). 4-bromofluorobenzene was used as the internal standard (IS), added to samples at fixed concentration of 100 ng mL -1 . The spiked "artificial" samples were processed exactly like the saliva samples (including incubation step, for 24 h). The Table 1 presents the used chemical standards and the concentrations evaluated as the quality control samples (QC). The limit of detection (LOD) was defined as the minimal concentration providing signal-to-noise ratio equal 3, the limit of quantitation (LOQ) was defined as the lowest concentration presenting imprecision and inaccuracy below 15%.

Precision and accuracy
The precision was evaluated: i) based in the samples analyzed in triplicate (3 aliquots); ii) model analytes spiked in concentrations corresponding to QC1 and QC3. The accuracy was assessed considering the relative standard error (= average experimental concentrationtheoretical concentration x 100 / theoretical concentration). Table 2 and Table 3 present the obtained results. The calculated imprecision and inaccuracy did not exceed 15% for the LOQ and 10% for other concentrations.

Stability
The stability of VOCs in samples was assessed for the following conditions: 24 h, at room temperature (21°C) and 2 thaw cycles (freezer at -20°C). The evaluation was made in triplicate, for two types of samples: i) processed pool of saliva samples, obtained from collected samples (n = 30), ii) spiked deionized water, at concentration levels corresponding to QC1 and QC3. The internal standard was added just prior analysis. For procedure i), stability was assessed calculating percentual variation of the response in comparison to samples freshly prepared and analyzed. Stability in approach ii) was described in terms of deviation from nominal concentration. The results are displayed in Table 4 and 5. The calculated data register alterations not superior to 15%, thus, sample stability can be attested.

Matrix effect
Matrix effect was evaluated in triplicate, using the response rate between spiked samples (0.5 mL of deionized water + standards) and solution of pure standards, both in concentrations corresponding to QC1 and QC3. To assess the extension of matrix effect, the normalized matrix factor (NMF) was calculated according to the Equation 1. Values near 1 indicate not considerable influence of the matrix in the response of the compounds, values much smaller than 1 indicate affinity between the chemical and the medium, while values above 1 suggest that the presence of the matrix enhances the partition of the substance to the gas phase. Finally, the relative standard error (RSD%) between values of NMF was calculated, to assure that the matrix effect is consistent and does not impair the conducted analysis with real samples. Results are presented in Table 6. Registered deviations were all below 15%.