Identification of Objective Metabolomics-based Biomarkers of Yogurt Consumption

Identification of Objective Metabolomics-based Biomarkers of Yogurt Consumption

Substantial evidence links yogurt consumption to a lower risk of type 2 diabetes (T2D). However, the existing evidence is derived exclusively from prospective cohort studies relying on self-reported dietary questionnaires, which-despite being validated-are subject to random and systematic errors that may compromise evidence quality and hinder regulatory approval.

The project aims to discover and validate biomarkers of yogurt intake.

The investigators will conduct a randomized, crossover, dose-response feeding trial involving 16 generally healthy adult participants (8 females, 8 males). The trial will include four 7-day diet periods, each separated by a 1-week washout. All diets will be based on a dairy-free background and supplemented with one of the following: (a) 1 serving of soy-based pudding (no yogurt), (b) 0.5 serving of yogurt, (c) 1 serving of yogurt, or (d) 2 servings of yogurt per 2,380 kcal/day.

On the 7th day of each diet, participants will complete a mixed-meal test at INAF, consuming a smoothie containing the same yogurt dose as during the preceding days. Plasma samples collected in the postprandial state will be used to profile over 20,000 metabolites. Using artificial intelligence-based approaches, potential biomarkers of yogurt intake will be identified. These candidates will then be filtered using standard statistical methods to assess dose- and time-responsiveness, robustness, and biological plausibility. Biomarkers that meet all criteria will be considered validated indicators of yogurt intake.

The U.S. Food and Drug Administration recently deemed the scientific evidence sufficiently robust to approve a petition for a qualified health claim on yogurt and reduced type 2 diabetes (T2D) risk. In contrast, Health Canada has yet to approve a similar claim. The existing evidence is derived exclusively from prospective cohort studies relying on self-reported dietary questionnaires, which-despite being validated-are subject to random and systematic errors that may compromise evidence quality and hinder regulatory approval.

Biomarkers of food intake are increasingly recognized as valuable tools to enhance objectivity in nutritional epidemiology. Metabolomics, the comprehensive profiling of plasma metabolites, is considered the gold standard for identifying such biomarkers. In the context of dairy, a major milestone was the identification of a metabolomic signature of yogurt intake in a large Spanish cohort. This signature-comprising 27 metabolites (e.g., C14:0 sphingomyelin, derived from myristic acid)-was predictive of yogurt intake in an independent U.S. cohort and associated with lower T2D incidence in both populations. However, because this signature was derived from observational data, it lacked specificity for yogurt intake.

This project seeks to fill this critical gap by discovering specific biomarkers of yogurt intake and validating them against recognized criteria.

The investigators will conduct a randomized, crossover, dose-response feeding trial involving 16 generally healthy adult participants (8 females, 8 males). The trial will consist of 4 predetermined diets, each lasting 7 days and separated by a 1-week washout period. On the 7th day of each diet, participants will undergo a mixed-meal test at INAF to collect plasma samples in postprandial state. The 4 diets will be based on a dairy-free background diet reflecting high adherence to Canada's food guide, with the following additions:

1. 125 g per 2,380 kcal per day of a soy-based pudding (i.e., yogurt-free diet);
2. 90 g per 2,380 kcal per day of yogurt (i.e., 0.5 reference serving of yogurt per 2,380 kcal per day);
3. 175 g per 2,380 kcal per day of yogurt (i.e., 1 reference serving of yogurt per 2,380 kcal per day);
4. 350 g per 2,380 kcal per day of yogurt (i.e., 2 reference servings per 2,380 kcal per day).

To ensure the identified biomarkers reflect the diversity of yogurts, four yogurt varieties will be tested:

a) fat-free plain regular yogurt; b) 2% milk fat plain regular yogurt; c) fat-free plain Greek yogurt; d) 2% milk fat plain Greek yogurt. Each yogurt type will be assigned to four participants, who will consume the same type consistently throughout the study. The goal is not to identify a distinct biomarker for each yogurt type but to ensure that the identified biomarker(s) account for the diversity of yogurt types. For the yogurt-free intervention, all participants will consume the same soy-based pudding. Participants will complete the four 7-day interventions in random order. This design ensures that any observed variations in the plasma metabolome will be attributed solely to yogurt.

During all four interventions, all meals and foods will be provided using a 7-day cyclic menu. The background diet for each intervention will be identical, dairy-free (i.e., excluding all dairy and butter), and designed to reflect high adherence to Canada's food guide, with an emphasis on minimally processed plant-based foods. To enhance acceptability and adherence, the menu will include at least one recipe from the official Canada's food guide documentation each day. A food technician will prepare the diets, with all foods and ingredients precisely weighed to the nearest 0.1 g.

In the context of intake biomarker identification, plain yogurt was deemed the most appropriate choice instead of flavored yogurt. However, plain yogurt is less palatable due to its sour taste. To address this, the background diet will include a daily serving of small fruits that participants will be instructed to either mix into the yogurt (or soy-pudding) or consume alongside. As the fruit serving will be standardized and consumed in consistent amounts each day across the four interventions, it will not introduce confounding in the yogurt biomarker identification process.

Across the four yogurt types and three levels of intake, yogurt will contribute an average of 120 kcal/day, as will the control soy-based pudding. The lowest caloric contribution will be 40 kcal (fat-free plain regular yogurt, 90 g/day), and the highest will be 240 kcal (2% milk fat plain Greek yogurt; 350 g/day). These calories will be added to a 2,380 kcal/day background diet to create a standardized menu that will average to 2,500 kcal/day across the four interventions. Given that the interventions last only seven days, it is unlikely that the difference in the caloric contribution of the yogurts or soy-based pudding will affect body weight or introduce confounding. The proposed approach-adding the tested foods to an identical isoenergetic background diet-is the optimal strategy. For a single participant, the only difference between the four diets will be the yogurt content, ensuring that any post-diet differences in the plasma metabolome are solely attributable to yogurt. Conversely, if the investigators opted to maintain a similar total caloric content of the background diet across the four phases, changes to the background diet would be required, which would introduce bias in the plasma metabolome. At baseline, participants' energy requirements will be estimated using validated predictive equations, and the total energy content of the diets will be adjusted accordingly to meet these requirements.

During each of the four 7-day interventions, participants will be required to consume at least three full meals at INAF under the clinical research coordinator (CRC)'s supervision. During these visits, they will also receive their meals for the next 24 to 72 hours. Participants will be instructed to consume only the provided foods and to consume all meals as supplied. Alcohol, vitamin supplements, and natural health products will not be permitted. Tea and coffee (≤2 cups/day) will be allowed, without daily fluctuations and without milk, cream or sugar. Unlimited amount of water will be allowed. Participants will be instructed to maintain their usual physical activity level stable for the entire protocol. On the 7th day of each diet, participants will undergo an acute mixed-meal test to inform on time-responsiveness of the yogurt biomarkers. After a 12-hour fast, participants will arrive at INAF at 7:30 AM. The test meal will consist of a breakfast smoothie containing the same yogurt (or soy-based pudding) provided daily during the preceding days. Plasma samples will be collected at baseline (fasting) and 1, 2, 4, and 6 hours post-meal. Samples will be drawn into EDTA tubes and stored at -80°C.

During the test, participants will be instructed to complete a validated self-administered web-based 24-hour dietary recall, specifically validated for research among Quebec adults. These data will be used to compare the performance of the R24W with that of the yogurt biomarkers in estimating the true yogurt content of the diet.

Participants and the CRC cannot be blinded to diet allocation. However, the exact amount of yogurt (or soy-based pudding) in the diets will not be disclosed to participants. The four diets will be randomly labeled with letters (e.g., A=high yogurt, B=soy, C=low yogurt, D=moderate yogurt). This approach will minimize information bias for the completion of the 24h recalls at the end of each intervention. Each intervention will be separated by a 1-week washout period, to provide participants with a break, reducing the risk of dropouts associated with a multi-week intervention. During the washout, participants will be instructed to return to their usual diet. The investigators do not anticipate any carryover effects of the diet during the washout on the subsequent 7-day intervention periods with respect to yogurt intake biomarker identification. Indeed, the identification of yogurt intake biomarker will leverage the plasma samples collected on day seven of each intervention, both in the fasting and postprandial states. Thus, it is highly unlikely that a biomarker from food consumed during the washout period exhibit the acute postprandial responsiveness expected for the yogurt intake biomarker under the proposed design. Total trial participation will be of 7 weeks. At the end of the trial, for each participant, the investigators will have a total of 15 plasma samples collected after yogurt consumption (i.e., 5 samples x 3 levels of yogurt intake) and 5 samples collected after the yogurt-free diet, for a total of 240 post-yogurt samples and 80 without yogurt (total=320).

Compliance and expected dropouts: A checklist will be provided to track daily food consumption. Having 3 full meals consumed at INAF under the CRC's supervision will reinforce adherence to the protocol.

Recruitment procedures: Inclusion criteria: age: 18-70 years; BMI: 18-40 kg/m²; willingness to consume plain yogurt daily for 3 weeks; for coffee/tea consumers, willingness to consume these beverages without added milk, cream or sugar for 4 weeks. Exclusion criteria: lactose intolerance; blood donation within 2 months; severe chronic illness or ongoing treatment for conditions such as infections, autoimmune diseases, cancer, type 2 diabetes, Crohn's disease, celiac disease, hepatic or renal diseases; history of severe cancer, cardiovascular disease or bariatric surgery; unstable body weight for 3 months; illicit drug use or alcoholism; tobacco/cannabis smoking/vaping (ensuring no day-to-day fluctuations and objective reporting of consumption is too complex); allergies or aversions to components of the diets.

Participants will be recruited from the Quebec City area through mass emails from INAF (outreach: >5,000) and Université Laval (>20,000). Interested individuals will be invited to contact the CRC for a screening phone call. Eligible individuals will then be invited to an in-person information session, where the full details of the protocol will be explained. Those interested will be asked to provide informed consent. After, participants will be randomized, and a baseline visit will be scheduled. A total of 10 different randomization sequences will be generated using William's design to ensure each treatment follows every other treatment an equal number of times. Each sequence will be attributed to a single participant, in a sex-specific fashion.

Metabolomics profiling: Once the trial is completed, plasma samples will be processed at INAF metabolomics platform using four complementary LC methods (positive and negative electrospray ionization) to determine the relative abundance of ~22,000 metabolites. These include amino acids, sugars, and small organic acids (BEH Amide); polyphenol metabolites and small peptides (HSS T3); free fatty acids and bile acids (CSH C18); and lipids (BEH C8). In addition to plasma samples, samples of each of the 4 yogurts and the soy pudding wil lbe processed.

Statistical analyses:

Step 1: identification of potential biomarkers of yogurt intake. To do this, the investigators aim to identify all metabolites allowing to distinguish between a yogurt-containing from a yogurt-free diet. Independent variables will include the relative abundance of all metabolites measured after each of the four interventions, considering plasma samples for each participant from the present trial (320 observations, of which 80 are yogurt-free). The dependent variable will be "presence of yogurt in the diet" (yes/no). The extreme gradient boosting algorithm (XGBoost) will be used to train classification models [42]. An iterative process will be repeated n times: (1) training the models, (2) identifying the metabolite with the best predictive ability, calculated by permutation testing, and (3) removing this metabolite before the next iteration. For each iteration, 25 models will be trained using different splits of participants into training (80%; n=484 observations) and testing (20%; n=120) sets. The iterations will continue until the sensitivity or specificity 95% confidence interval of the models reaches 0.5 (null predictive power of a classification algorithm). At the end of this step, the n removed metabolites will constitute the pool of potential biomarkers of yogurt intake, which will be retained for subsequent steps. The subsequent steps will leverage standard statistical approaches to test the key characteristics of a validated biomarker of food intake, as proposed by FoodBALL. Unless otherwise specified, all the following tests will be conducted using mixed models for repeated measures, with metabolite relative abundance as the dependent variable. The models will be adjusted for age, sex, body mass index, and total daily calories from the diet.

Step 2, crude dose-responsiveness: The investigators will compare the relative abundance of each potential biomarker following yogurt-free and yogurt-containing diets. Only metabolites whose abundance increases with yogurt consumption will proceed to the next step.

Step 3, refined dose-responsiveness: The investigators will compare the relative abundance of metabolites across the three levels of yogurt intake tested in the study. Metabolites showing a significant linear trend, with the yogurt-free diet coded as 0, will be retained.

Step 4, time-responsiveness: Using data from plasma samples collected during the acute test meal, the investigators will assess whether the relative abundance of metabolites increases over time. Metabolites with a greater area under the curve (AUC) following yogurt intake compared to the soy-based pudding will be retained.

Step 5, plausibility: The investigators will assess the plausibility of the retained metabolites as originating from yogurt based on their nature and chemical structures, using the Milk Composition Database. In addition, yhe investigators will compare these metabolites to the metabolomics profiles of the four tested yogurts. After this step, the retained metabolites will have already satisfied FoodBALL's essential criteria for being considered validated biomarkers of yogurt intake.

Inclusion criteria:

• age: 18-70 years;
• BMI: 18-40 kg/m²;
• willingness to consume plain yogurt daily for 3 weeks;
• for coffee/tea consumers, willingness to consume these beverages without added milk, cream or sugar for 4 weeks.

Exclusion criteria:

• lactose intolerance; blood donation within 2 months; severe chronic illness or ongoing treatment for conditions such as infections, autoimmune diseases, cancer, type 2 diabetes, Crohn's disease, celiac disease, hepatic or renal diseases;
• history of severe cancer, cardiovascular disease or bariatric surgery;
• unstable body weight for 3 months;
• illicit drug use or alcoholism; tobacco/cannabis smoking/vaping (ensuring no day-to-day fluctuations and objective reporting of consumption is too complex);
• allergies or aversions to components of the diets.