**1. Introduction**

Currently, the understanding of diet-health relationships has gradually shifted from individual dietary components to overall dietary patterns that beneficially modulate metabolic physiology [1]. In this regard, several epidemiological and clinical studies have shown that the traditional Mediterranean-style eating pattern (characterized by high intake of fruits and vegetables, olive oil, legumes, whole grains, and fish and moderate consumption of white meat, dairy and wine during meals) has many health benefits [2,3], including beneficial changes in biomarkers of CVD risk [4,5] and lower risk of cardiovascular disease (CVD) [6–8].

However, the exact mechanisms of the benefits of the Mediterranean-style dietary pattern have ye<sup>t</sup> to be understood. The application of metabolomics to measure changes in biological variables in response to dietary interventions has been proposed as a potential tool to discover biomarkers associated with healthier eating [9–13]. There is observational epidemiological evidence that acylcarnitines, Trimethylamine N-oxide (TMAO), some amino acids such as phenylalanine, glutamate as well as several lipid classes are associated with CVD risk [14]. While the effects of individual dietary ingredients on metabolome have been reported [15–20], only a few studies have focused on overall dietary patterns [21–26]. Some cross-sectional studies such as the Whitehall II study showed that a healthy diet was associated with specific fatty acids that reduced the risk of CVD [27]. In a British population, the association between the adherence to the MedDiet and cardiometabolic outcomes was mediated by baseline levels of acylcarnitines, sphingolipids, and phospholipids [24]. In the Supplementation en Vitamines et Mineraux Antioxydants (SU.VI.MAX) study, some metabolites were also cross-sectionally associated with dietary recommendations [28]. More recently, a metabolic signature of the Mediterranean Diet has been consistently identified in two large cohorts [26].

Importantly, studies analyzing changes in metabolites levels in response to MedDiet interventions are still scarce and inconclusive. In the Metabolic Syndrome Reduction in Navarra (RESMENA)study, after 2 months of an energy restricted MedDiet intervention, some significant changes in metabolites were shown but they were no longer observed after 6 months [22]. Additional evidence comes from the Prevention with Mediterranean Diet (PREDIMED) study, where several a priori-designed analyses found that the MedDiet may reduce the deleterious effect on cardiovascular or type-2 diabetes risk associated with 1 year changes in branched-chain amino acids [29,30], carnitines [31] and other metabolites. However, in most of these analyses, the 1 year metabolite changes were similar between the intervention and the control group, suggesting that the observed protective effect of the MedDiet could be due to other mechanisms.

Therefore, the objective of this study is to identify changes in plasma metabolic biomarkers associated with a MedDiet intervention within a subsample of a cluster-randomized controlled trial (Feeding America's Bravest) among firefighters. We hypothesize that the MedDiet intervention induces changes in metabolites within clinically relevant pathways.

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

#### *2.1. Study Design and Participants*

The overall study design, intervention strategies and primary outcomes of the Feeding America's Bravest trial have been previously reported [32]. Briefly, Feeding America's Bravest is a step-wedge cluster-randomized controlled trial within the 44 stations of Indianapolis Fire Department, which aims to compare a MedDiet Nutritional Intervention vs. an ad libitum Midwestern-style diet (control or no intervention) during 12 months. Group 1 (or the intervention group for 12 months) continued under a self-sustained continuation phase for another 12 months. Group 2 (initially controls) crossed over (at 12 months) to receive the active Mediterranean Diet Nutritional Intervention for 6 months followed by another 6 months of a self-sustained intervention phase. For this nested study, we randomly selected a sub-group of participants (*n* = 48) whose fire stations had been assigned to the MedDiet intervention (*n* = 24) or the control group (*n* = 24) for the previous 12 months. At that time (time point 1 for our study and 12 months follow up of the parent study), the intervention group (Group 1) continued under a self-sustained phase for another 6 months (time point 2 for our study or 18 months of the parent study) and the control group underwent the active MedDiet intervention for 6 months. Plasma metabolic biomarkers were analyzed at the two time points (Figure 1). The overarching Feeding America's Bravest protocol was approved by the Harvard Institutional Review Board (IRB16-0170) and is registered at Clinical Trials (NCT02941757). All participants provided informed consent for participation.

**Figure 1.** Study design and timeline of the metabolomics study within Feeding America´s Bravest parent study (step-wedge cluster-randomized control trial).

#### *2.2. Mediterranean Diet Intervention*

At the two time points of this study, data on sociodemographic characteristics, physical activity, sleep behaviors (a modified Pittsburgh Sleep Quality Index, where on a scale from 0 to 7 participants were asked to identify the statement which option best described their habitual level of physical activity over the past month [33]), food consumption based on a self- reported validated food frequency questionnaire [34] and a modified Mediterranean Diet scale [35], and anthropometric and clinical variables were assessed.

The MedDiet intervention has been described in detail elsewhere [32]. Briefly, it consisted of educational sessions and videos created by a certified nutritionist, leaflet and recommendations about the Mediterranean Dietand lifestyle, firefighter-tailored Mediterranean recipes (by modifying Firefighters´ favorite recipes according to the MedDiet principles by a chef and a nutritionist), in-site chef cooking demonstration, a firefighters' food pyramid, Mediterranean food samples and discount coupons to a large supermarket chain for specific Mediterranean Diet-compatible foods [32].

Adherence to the MedDiet was assessed by the validated modified Mediterranean Diet Score (mMDS) [35,36] and the PREDIMED score. (36) The mMDS range from 0 (lowest) to 51 (highest conformity to the MedDiet) [36] and consists of 13 domains including consumption of fast food, fruits, vegetables, legumes, nuts, sweet desserts, fried foods, ocean fish, breads and starches (consumed at home and the fire station), the type and frequency of alcoholic beverages, non-alcoholic beverages (consumed at home and the fire station) and the type of cooking oil or fat (consumed at home and the fire station). Each component ranged from 0 (less adherence) to 4 points (better adherence) except the type of alcohol (wine; 0–2 points) and the type of cooking oil or fat (0–5 points). Weighted scores was considered for domains evaluated at both the homes and fire stations based on the percentage of meals consumed at each location [36].

The PREDIMED score [37] consists of 14 questions; 12 of them are about food consumption frequency (olive oil consumption, vegetables, fruits, red/processed meats, butter/margarine, soda drinks, wine, legumes, fish/seafood, nuts, commercial sweets, sofrito consumption), and another two about food intake habits considered characteristic of the Spanish Mediterranean Diet (preference of poultry consumption instead of red meats, use of olive oil as main culinary fat). Each question was scored 0 or 1, with a total possible range of 0 to 14; higher scores indicate greater adherence to the MedDiet.

#### *2.3. Collection of Plasma Samples*

In this nested study, 12 h fasting blood samples were collected at time point 1 (baseline for this metabolomics study and 12 months for the trial) and time point 2 (6 months follow up for this metabolomics study and 18 months for the trial) of follow up; samples were kept cold and immediately processed to separate the plasma with a refrigerated centrifuge. Next, the 200 μL cryovials were placed at −80 ◦C.

#### *2.4. Plasma Biomarkers Measurements*

Metabolites were quantified in plasma samples from 83 individuals that had optimal values using high-throughput proton Nuclear Magnetic Resonance (NMR) metabolomics (Nightingale Health Ltd., Helsinki, Finland). This method provides simultaneous quantification of routine lipids, lipoprotein subclass profiling with lipid concentrations within 14 subclasses, fatty acid composition, and various low-molecular metabolites including amino acids, ketone bodies and glycolysis-related metabolites. Details of the experimentation and applications of the NMR metabolomics platform have been described previously [38]. All measured metabolites fall in the range of detection; representative coefficients of variations (CVs) for the metabolic biomarkers were published previously [39,40].
