We identified saliva, plasma, and multi-fluid signatures of adiposity measures. Metabolites included those in the metabolism pathways of amino acids (alanine, aspartate, tryptophan, proline, arginine, polyamine, tyrosine, and lysine), nucleotides (purines and pyrimidines), carbohydrates (fructose, mannose, and galactose), and lipids (glycerolipid and fatty acid). Saliva, but not plasma, signatures of BMI were associated with a higher risk of any type of diabetes progression. On the other hand, saliva, and multi-fluid, but not plasma, signatures were associated with a higher risk of progression from prediabetes to T2D after accounting for differences in anthropometric measures of BMI and WC. The associations between metabolomic signatures and diabetes progression were magnified after further adjusting for BMI and WC. A possible explanation is that anthropometric measures of BMI and WC alone may not fully capture the intrinsic variation in the metabolism that is associated with diabetes progression. Our metabolomic signatures, on the other hand, were able to capture this additional information. Another possible explanation is that variations in the metabolomic signatures of adiposity measures are independent of BMI or WC alone. When BMI and WC are held constant, there is still enough variation to predict T2D risk, which indicates that the metabolomic signatures are not on the causal pathway between adiposity measures and T2D at year 3.
Our observations align with results from other studies that identified metabolomic profiles for overweight and obesity. A systematic review of published articles on metabolomic profiles (primarily from plasma) of obesity summarized that branched-chain amino acids (BCAA) (leucine, isoleucine, and valine) and aromatic amino acids (phenylalanine, tyrosine, tryptophan, and methionine) are positively associated with higher BMI and WC. On the other hand, glutamine and glycine levels are typically lower in individuals with obesity compared to their non-obese counterparts21. We identified three positively weighted plasma metabolites in the pathway of BCAAs in the profiles of BMI or WC: alpha-hydroxyisovalerate, alpha-hydroxyisocaproate, and N-acetylleucine, although we also identified several metabolites in this pathway that are negatively weighted. Several plasma metabolites in pathways of aromatic amino acids were also selected in our profiles for both BMI and WC, although some were positively weighted while others were negatively weighted. We also identified glutamine and glycine to be negatively weighted in our multi-fluid and plasma profiles for WC, which confirms findings from previous literature.
Some other highly weighted metabolites in amino acid pathways have also been reported to be associated with obesity in other literature. Cystine, the disulfide form of cysteine, was highly positively weighted in our plasma profile for WC. In fact, multiple epidemiology studies have provided evidence that links cysteine to obesity22. Evidence suggests that high levels of total plasma cysteine predict obesity and are independently associated with BMI, fat mass, and insulin resistance23–25. As a well-established biomarker for obesity, some researchers targeted lowering plasma total cysteine levels when testing drugs for obesity control26. While there is not much evidence linking cysteine, the sulfur-containing form of cystine, to obesity, cystine itself has been reported to enhance adipogenesis and lipid accumulation in human preadipocytes and is positively associated with fat mass27. Hydroxyasparagine, the metabolite with the highest positive weight in our plasma and multi-fluid signatures for BMI, is found in the prenatal metabolomic profiles of mothers of obese children28. In another study among Arabs, high levels of hydroxyasparagine were found in individuals with mild-obesity-related diabetes29. Kynurenine, a metabolite identified in our plasma profile for BMI, is converted from tryptophan in the brain and the periphery and affects various biological actions, including modulating immune response and inflammation30. The tryptophan-kynurenine pathway has been identified as a novel target for preventive and therapeutic interventions in diabetes. Substantial evidence provided by animal, human, and microbiome studies suggests that degraded conversion from tryptophan to kynurenines plays a role in diabetes development and insulin resistance31–33. In addition, it was found to positively correlate with both BMI and WC in a cohort of American Indians34.
Among carbohydrate metabolites, several studies have reported that increases in mannose and glycerol levels are seen in the plasma of obese men35, which were both found in our plasma profile for BMI with positive weights. As for lipid metabolites, studies have observed altered levels of choline in obese participants, although the direction of association varies by study21. In our results, plasma choline had positive weights in both the plasma profile of WC and the multi-fluid profile of BMI. Several sphingomyelins were selected in the plasma and multi-fluid profiles for both BMI and WC. Sphingomyelins have been found to be associated with T2D progression in a group of pre-diabetic Korean men36 and a meta-analysis of 71,196 participants37. Citrate, a metabolite in the TCA cycle, has been found in low levels in obese children38 and is also one of the negatively weighted metabolites in our plasma profile for BMI.
Although many of our findings aligned well with previous studies, there are also some discrepancies between our results and previous studies. One study found levels of alanine to be higher in obese participants compared with the non-obese individuals39. In our analysis, alanine was negatively weighted in the plasma profile for WC. Another study done in a cohort of Chinese adults reported that betaine was negatively correlated with WC40, although in our study it was positively weighted in the plasma metabolomic profile for WC. Potential reasons for these discrepancies could be due to differences in study population.
Some highly positively weighted saliva metabolites selected for both saliva and plasma profiles include 2R,3R-dihydroxybutyate, kynurenate, urate, and cis-3,4-methyleneheptanoylcarnitine. Although no literature has reported on salivary 2R,3R-dihydroxybutyrate, its urinary form has been found to be positively correlated with BMI in a group of ethnically diverse pregnant women41. Salivary kynurenate is the conjugate base of kynurenic acid, which has been associated with dental abscesses, psychological stress, schizophrenia, and glioblastoma42–45. It is a metabolite of kynurenine, whose association with T2D has been described above. Urate, also called uric acid, is the end product in the breakdown process of the nucleotide purine. Elevated levels of salivary urate are associated with cancer, HIV infection, gout, hypertension, and emotional disorders, whereas lower levels are associated with Alzheimer’s disease, multiple sclerosis progression, and cognitive impairment46–49. It has also been reported that levels of urate are significantly higher in saliva samples of individuals with diabetes50,51, especially among those with uncontrolled diabetes52. Urinary cis-3,4-methyleneheptanoylcarnitine has been reported to be associated with inherited metabolic diseases and weight loss53,54, although there is no data on this metabolite in human saliva.
Our findings should be interpreted in the context of a few limitations. First, given that all participants were overweight or obese at baseline, the study population is considered high-risk. As a result, the rate of progression from normal to pre-diabetes to T2D may be quick and may not be entirely linear. This also limits the generalizability of our observations. Because we did not have access to metabolomic profiles of people in the underweight or normal weight category, the metabolomic profiles we identified are not reflective of the complete ranges of BMI and WC. Second, due to the high-dimensional nature of the metabolomics data and the relatively modest sample size, we were limited in statistical power. In addition, because this is the first large cohort with available saliva metabolomics data, we were not able to validate our results in an external testing set. Although we split participants into a training and a testing set, there is still potential for overfitting. Despite these limitations, our study also has several strengths. This is the first study to identify saliva, plasma, and multi-fluid metabolomic signatures of adiposity measures among Puerto Ricans and examine their association with diabetes progression and periodontal disease progression. We leveraged a large number of plasma and saliva metabolites among Puerto Rican adults who have established health disparities and have a high prevalence of obesity and T2D. The longitudinal design and repeated laboratory assessment of diabetes in the SOALS cohort allowed us to examine differences in the associations of the signatures with different stages of disease progression.
In conclusion, our study found that saliva metabolomic signatures were more predictive of diabetes progression over three years among overweight and obese Puerto Ricans. Given that saliva samples are non-invasive and more accessible than plasma samples, utilizing salivary metabolites for the prognosis of T2D in a clinical or health care setting would make screening faster and cost-effective and can potentially lower health disparities. Future studies should try to replicate these findings in other populations.