The prevalence of pediatric obesity has increased worldwide in recent years and raised urgent concern about expected metabolic dysregulation and serious comorbidities, such as T2D, likely to arise among obese children as they age into adulthood [1]. The relationship is unknown among adult-onset, obesity-associated pathologies and exposure of these individuals to diet-induced obesity while still children. Furthermore, it is not known if at-risk cohorts of pediatric subjects can be identified by non-invasive measures early enough during development of pediatric obesity for effective interventions to reduce risk of T2D. The importance of the answers to these questions is widely recognized, thus countries with significant prevalence of pediatric obesity are devoting resources to understand the scope of this problem and devise appropriate public health recommendations.

In this research, we describe an approach that uses saliva samples as a basis for diagnosis. The selection of biomarkers for evaluation in this study of obese children was focused on adipokines (adipocytokines), because concentrations of these factors partly define metabolic health. A subset of these proteins is of particular interest for immunometabolism, including adiponectin, leptin and resistin, which are adipocyte-derived hormones secreted in saliva [2], salivary [3] monocyte chemotactic protein-1 (MCP-1) and salivary [4], [5] tumor necrosis factor-alpha (TNF-α). Interleukins (IL) comprise a second major class of investigated biomarkers, each of which has also been measured in saliva, but not in an obese adolescent population. Specifically, IL-1β is associated with periodontal inflammation [6]; IL-6 has been measured in periodontitis patients; IL-4, IL-10, IL-12 and IL-17 are related to Sjogren’s syndrome [7]; IL-10 is reduced in periodontitis patients [5]; IL-8 is related to dental caries in adolescents [4] and oral cancer [8]; IL-13 is identified in the sputum of asthmatics [9]; IL-17 is lower in patients with periodontal disease [10]; and interferon γ (IFN-γ) is higher in the saliva of control subjects without Sjogren’s syndrome [7]. Thus, adipokines and cytokines from sources other than peripheral blood have been assayed and associated with disease states. Both salivary immunoreactive insulin [11] and salivary ghrelin [12] have been associated with T2D and obesity, and were included in our analysis. Two oral inflammatory disease markers were also included, matrix metalloproteinase-9 (MMP-9) and myeloperoxidase (MPO). MMP-9 is a protease often found elevated in patients with periodontal disease [13] and oral cancer [14]; whereas MPO, a peroxidase often used as a measure of neutrophil degranulation, is elevated in diabetic patients [15]. Vascular endothelial growth factor (VEGF) was included because it appears in saliva of diabetic women [16]. C-reactive protein (CRP) was included because it is associated with inflammation and cardiovascular disease [17]. Collectively, the panel defines inflammatory and metabolic processes important for metabolic health in obesity.

Blood elements that partition into saliva have attracted interest [18]. Saliva has major advantages over peripheral blood for the study of disease development in childhood because sample collection is non-invasive. Here, we collected 8,319 saliva samples from 10–12-year old Kuwaiti children, a high risk population for obesity and T2D development [19]. We measured salivary concentrations of 20 hormones and cytokines in a random subset of 744 subjects. The well-established relationship between insulin-resistant obesity and systemic elevation of pro-inflammatory cytokines [20] supports our hypothesis that the saliva of obese children shows informative imbalances in critical immunometabolic factors [21]. The results of this study provide insight into the early development of metabolic disease in children and establish that non-invasive methods are robust and useful for data collection in studies of vulnerable subjects.

Saliva samples (744) were randomly selected from the 8,319 total to provide 186 for each category of underweight, normal healthy weight, overweight and obese. Characteristics of these groups are evaluated in Table 1. No significant differences in age between groups were noted. BMI, WC, blood pressure and fitness exhibited significant incremental increases as the body weight category moved to obese. In every body weight category, BMI and WC significantly differed from each other. Systolic and diastolic blood pressures were significantly different only between normal and overweight or obese subjects. Obese subjects had significantly lower fitness as measured by exercise stimulated heart rate elevation than any other category. Obese subjects had 64% higher BMI, 40% greater WC, 17% higher systolic and diastolic blood pressure and 50% greater exercise elevation of heart rate as a measure of unfitness compared to normal weight subjects. Table S1 in File S1 shows similar statistics from 53 US subjects used to measure the relative concentration of insulin in saliva and plasma.

The assay performance is summarized in Table S2 in File S1. The median concentration of 17 biomarkers was greater than the manufacturer’s stated assay sensitivity. Median values fell below the assay sensitivity for IL-10, leptin and ghrelin. The analysis software allowed extrapolation beyond the lowest standard, provided the value did not fall on or below the blank. Using this feature, although the concentration cannot be said to have been accurately measured, the assay provided a means to order low concentrations for non-parametric rank analysis. All probes provided non-zero values for 95 to 100% of the samples tested except for leptin; 37.4% of the leptin determinations were too low to measure. In the leptin analysis, however, the percentage of measurable samples increased with increasing obesity so that analysis by non-parametric rank could be performed without introducing bias.

The assay results are summarized in Figure 1 and Table 2. Of the 20 probes tested, significant differences between obese and normal weight subjects occurred in the concentration of insulin, CRP, adiponectin and leptin. CRP and insulin were significantly elevated in obese relative to normal subjects. Salivary insulin levels in obese subjects (median = 127 pg/ml) were almost 3 times that of normal weight subjects (median = 44 pg/ml) and salivary CRP of obese subjects (median = 435 pg/ml) was almost 6 times that of normal weight subjects (median = 76 pg/ml). The estimated concentration of salivary leptin in obese subjects (median = 3.3 pg/ml) was 3 times that of normal weight subjects (1 pg/ml). Salivary adiponectin decreased by approximately 30% with increasing obesity from that of normal healthy weight subjects (median 4,083 pg/ml) to overweight and obese subjects (median = 2981 and 2798 pg/ml respectively). Differences between normal and obese subjects for insulin, CRP, adiponectin and leptin were all highly significant (p<0.0001). There were no significant differences between biomarker concentrations of underweight and normal healthy weight subjects. Salivary levels of the remaining 16 probes (Figure S1, Table S3 in File S1) did not significantly change when analyzed by body weight categories.

Evaluation of saliva supernatants normalized by salivary total protein content (Table 3), indicated that significant levels of biomarkers obtained were similar to those measured using salivary concentration. Salivary protein concentration significantly decreased with increasing salivary flow rate (Figure 2). Neither salivary protein concentration nor salivary flow rate varied significantly between body weight categories (Figure 3).

The diagnostic implications of these findings were evaluated by classification tree topology (Figure 4A). Of the obese subjects, 76.3% had >219 pg/ml CRP in their saliva, indicating that the inflammatory state was the most common form of obesity. Of the obese subjects with lower levels of CRP (<219 pg/ml), 13% had high insulin (>128 pg/ml) and 11% had low insulin. The obese group with CRP <219 and insulin <128 would be predicted to be ‘metabolically healthy’. Using only salivary CRP and insulin as predictors, the overall diagnostic sensitivity for identifying obesity by biomarker status was 89% and specificity was 61%.

We used Luminex analysis to evaluate correspondence between saliva and plasma concentration of insulin in a group of 53 American subjects similar in age to the Kuwaiti subjects (Figure 4B). We found that the concentration of immunoreactive saliva insulin is approximately half of the plasma concentration and estimates of plasma insulin with high correlation (r² = 0.67). By this approximation, the predictor variable of 128 pg/ml salivary insulin is equivalent to 68 pmoles/L = 11 µU/ml in plasma, a value higher than that reported for normal fasting subjects 38–46 pmoles/L = 6.3–7.6 µU/ml [27].

Investigation of the properties of each of the six groups described by this categorization tree revealed significant differences relative to the normal healthy (NH) group in both clinical measures and biomarker concentrations (Table 4). Systolic blood pressure, BMI, WC and insulin were significantly elevated in all categories except for the non-obese healthy group (NH). Diastolic blood pressure was significantly elevated in the obese with high insulin (OI) and obese with high CRP (OC) groups. Low fitness was significantly increased in the OC group. CRP and IL-6 were significantly elevated in both high CRP groups (OC and NC) and leptin was significantly elevated in the obese high CRP group (OC). The non-obese, high CRP group (NC) exhibited elevation of multiple biomarkers (IL-10, resistin, IL-1β and MMP-9). The obese but healthy group (OH) exhibited significantly lower concentrations of IL-10 and adiponectin.

Many blood factors also occur in saliva and their levels often correlate. In our study, we surveyed 20 possible biomarkers related to obesity and found four that exhibit significant change with increasing body weight in a pediatric population. These data suggest that saliva could be a useful blood surrogate for the study of metabolic complications of obesity in children, where repeated blood sampling can be both traumatic and difficult.

Elevated plasma insulin is characteristic of T2D and also is proportional to body fat content [28]. In agreement with our data, investigators have noted that salivary insulin levels in obese children may be four or five times higher than in normal healthy weight children [29] and that salivary insulin correlates well with plasma levels at only slightly lower concentration [30]. Salivary insulin is linearly related to plasma insulin during the glucose tolerance test [11] and correlates well with plasma concentration after insulin injection [31]. Plasma insulin decreases in parallel with weight loss in obese children [32].

Salivary CRP concentration has also been found to correlate well with serum concentrations [24]. Association between salivary CRP and obesity has been reported in black South African children [33]. Human studies associate high levels of CRP with metabolic syndrome and T2D [34]. Reduced adiponectin is an independent risk factor for metabolic syndrome [35]. Both salivary adiponectin [24] and leptin [2] concentrations are highly correlated with plasma concentrations.

Since it is well-known that the concentration of proteins can vary significantly in response to stimulation, total protein is sometimes used to normalize concentration of specific salivary analytes [36]. The total protein content of salivary samples was reduced as salivary flow rate increased (Figure 2), however, neither total salivary protein nor salivary flow rate were significantly different between body weight categories (Figure 3) indicating that obesity did not affect salivary protein levels or salivary flow rate in this population. Investigation of a protein-normalized sample subset indicated that this procedure did not appreciably modify the diagnostic potential of salivary insulin, CRP, leptin or adiponectin in the samples collected from adolescent Kuwaiti children (Table 3).

The first level of the classification tree (CRP >219 pg/ml, Figure 4A), is associated with multiple inflammatory mediators (Table 4). The obese high salivary CRP (OC), represented the largest group of obese subjects (76.3%). The major differences among groups were elevated diastolic blood pressure, significantly reduced fitness and elevated leptin in the obese (OC) group. The non-obese high salivary CRP group (NC) had high levels of CRP, insulin, IL-6, IL-10, resistin, IL-1β and MMP-9 compared to the normal healthy group (NH).

The second level of the classification tree (CRP ≤219, insulin >128 pg/ml) also identified two groups of subjects (OI and NI), but these did not have elevated salivary inflammatory mediators but did have elevated insulin levels. The principal differences were that obese subjects had high systolic and diastolic blood pressures. Subjects with both low insulin and low CRP included a normal healthy group (NH) and an obese, ‘healthy’ group (OH). In this case, obese subjects were associated with elevated insulin, reduced adiponectin and reduced IL-10.

Comparison of these observations with those based on immunohistochemistry of obese adult adipose tissue [37] strongly suggests that the group designated as CLS+ (macrophage crown-like structure positive), found in 72% of obese adults, may coincide with the subjects designated as CRP >219 in our studies (76.3% of obese subjects). Our studies suggest, however, that non-obese cohorts with comparable biomarker levels could be at risk for development of immunometabolic complications of obesity.

The constellation of biomarkers that appear in saliva are likely a reflection of underlying pathology [38]–[40]. High concentrations of salivary CRP, myoglobin and MPO, for example appear after myocardial infarction [41]. Periodontal disease characteristically exhibits high levels of salivary IL-8 [42], MMP-9 and IL-1β [43]. Inflammatory bowel disease is associated with elevated salivary IL-6 and CRP [44]. We can now conclude that adolescent obesity is most commonly associated with high levels of salivary insulin and CRP.

We have identified four salivary biomarkers in 10-year old subjects that change significantly with increasing obesity; insulin, CRP, adiponectin and leptin. The results of this study suggest that obesity may be characterized and classified by salivary biomarker concentrations. Use of these relatively non-invasive markers, particularly in longitudinal studies, to investigate development of metabolic diseases in children and evaluate therapeutic interventions could be used to evaluate preventative therapies.