Abstract It is well recognized that patients with severe obesity exhibit remarkable heterogeneity in response to different types of weight-loss interventions. Those who undergo Roux-en-Y gastric bypass (RYGB) usually exhibit more favorable glycemic outcomes than those who receive adjustable gastric banding (BAND) or intensive medical intervention (IMI). The molecular mechanisms behind these observations, however, remain largely unknown. To identify the plasma metabolites associated with differential glycemic outcomes induced by weight-loss intervention, we studied 75 patients with severe obesity (25 each in RYGB, BAND, or IMI). Using untargeted metabolomics, we repeatedly measured 364 metabolites in plasma samples at baseline and 1-year after intervention. Linear regression was used to examine whether baseline metabolites or changes in metabolites are associated with differential glycemic outcomes in response to different types of weight-loss intervention, adjusting for sex, baseline age, and BMI as well as weight loss. Network analyses were performed to identify differential metabolic pathways involved in the observed associations. After correction for multiple testing (q < 0.05), 33 (RYGB vs. IMI) and 28 (RYGB vs. BAND) baseline metabolites were associated with changes in fasting plasma glucose (FPG) or glycated hemoglobin (HbA1c). Longitudinal changes in 38 (RYGB vs. IMI) and 38 metabolites (RYGB vs. BAND) were significantly associated with changes in FPG or HbA1c. The identified metabolites are enriched in pathways involved in the biosynthesis of aminoacyl-tRNA and branched-chain amino acids. Weight-loss intervention evokes extensive changes in plasma metabolites, and the altered metabolome may underlie the differential glycemic outcomes in response to different types of weight-loss intervention, independent of weight loss itself. Keywords: obesity, bariatric surgery, metabolomics, glycemic outcomes 1. Introduction Morbid obesity, defined as a body mass index (BMI) ≥ 40 kg/m^2 or BMI ≥ 35 kg/m^2 with comorbidities (e.g., diabetes), is a serious health problem that impedes normal daily life and activities [[44]1]. Patients with morbid obesity suffer from an increased risk of type-2 diabetes [[45]2], hypertension [[46]3], heart disease [[47]4], and cancer [[48]4]. Although non-surgical interventions, such as calorie restriction, physical activity, and medication management, are effective in achieving short-term weight loss [[49]5], bariatric surgery has proven to be most effective in achieving sustainable weight loss and long-term beneficial glycemic outcomes [[50]6,[51]7]. Of the several bariatric surgical procedures, such as adjustable gastric banding (BAND) and Roux-en-Y gastric bypass (RYGB), RYGB typically results in more weight loss and a higher rate of diabetes remission [[52]8,[53]9]. However, patients undergoing bariatric surgery exhibit remarkable heterogeneity in response to different types of surgical procedures as well as post-operative outcomes. Although several mechanisms such as genetics [[54]10], epigenetics [[55]11], and gut hormone [[56]12] have been postulated to explain the beneficial effects of bariatric surgery, the precise mechanisms through which bariatric surgery induces favorable and heterogeneous glycemic outcomes remain largely unknown and unexplored. A better understanding of such mechanisms may lead to the discovery of novel mechanistic markers that can be used to identify patients who may benefit from a specific type of weight-loss intervention, thereby achieving the precise treatment of obesity. Metabolomics is an emerging high-throughput biochemical technique that can identify and quantify numerous small metabolites (<1500 Da) in a biological sample. There is evidence that bariatric surgery can rewire cellular metabolism and results in extensive metabolic changes [[57]13,[58]14]. As an individual’s metabolic profile provides the functional readout of his/her current metabolic state that is responsive to both intrinsic and extrinsic perturbations [[59]15,[60]16], metabolomics provides a powerful tool to characterize the entire “metabolome” (all metabolites in a biological sample) and discover previously undescribed metabolic disturbances evoked by weight-loss intervention. In fact, changes in several metabolites such as branched-chain amino acids (BCAAs) [[61]17,[62]18], glycerophospholipids [[63]13,[64]19], and bile acids [[65]20], have been shown to be differentially responsive to surgical versus lifestyle/behavioral interventions. However, several research gaps exist in this field. First, previous metabolomic studies on bariatric surgery were largely cross-sectional, which did not examine the relationship between a change in metabolites and a change in glycemic outcomes. Second, most previous studies employed a targeted approach by focusing on a preselected list of metabolites, resulting in a low coverage of the metabolome [[66]21,[67]22]. Third, to the best of our knowledge, very few previous metabolomic studies on bariatric surgery examined whether alterations in the blood metabolome underlie the differential glycemic outcomes in response to different types of weight-loss intervention. Fourth, most existing metabolomic studies on bariatric surgery did not control for weight loss, making it hard to interpret whether the observed post-operative metabolic improvements are due to bariatric surgery itself or subsequent weight loss. Our hypothesis is that metabolomic alterations underline the differential glycemic outcomes in response to different types of weight-loss intervention, independent of weight loss. Disentangling the weight-loss-independent mechanism is important because recent evidence [[68]23,[69]24] suggests that there are additional important mechanisms through which bariatric surgery can result in the improvement of glycemic outcomes, independent of weight loss. Using a longitudinal study design and an untargeted metabolomic approach, the goals of this study are to (1) identify baseline plasma metabolites associated with the differential glycemic outcomes in response to different types of weight-loss interventions, independent of known clinical factors and weight loss; and (2) examine the longitudinal association between a change in plasma metabolome and a change in glycemic indices (before and after weight-loss intervention), independent of baseline clinical factors and weight loss. 2. Materials and Methods 2.1. Study Participants The HeadsUp Study (July 2011–June 2016), funded by the State of Louisiana, was a longitudinal study designed for obesity management among patients with obesity (aged 21–70 years, 83% women, 29% black) in the State of Louisiana Office of Group Benefits (OGB). A detailed description of the study design and the inclusion and exclusion criteria has been previously reported [[70]25]. Briefly, the HeadsUp Study was designed to examine the effectiveness of bariatric surgery and non-surgical intervention on short-term (6 months) and long-term (≥1 year) health outcomes, thereby providing guidance for future treatment options for patients with severe obesity [[71]25]. Eligibility criteria for the surgical intervention included BMI > 40 kg/m^2 or BMI > 35 kg/m^2 with type-2 diabetes, and BMI ≥ 33 kg/m^2 to be eligible for the intensive medical intervention (IMI) program. Exclusion criteria of the HeadsUp Study included (1) individuals who failed to complete the initial web-based or follow-up telephone screenings; (2) individuals who did not meet the BMI thresholds (BMI < 33 kg/m^2 for non-surgical interventions and BMI < 35 kg/m^2 without type-2 diabetes or <40 kg/m^2 for surgical interventions); (3) individuals who did not achieve the required pre-screening weight loss of at least 4 pounds; (4) women who were pregnant or intended to become pregnant within 3 years; (5) individuals with significant medical conditions or psychiatric disorders; and (6) individuals who underwent prior bariatric surgeries or those who were unable to comply with the study protocol. A total of 75 participants were enrolled into three equal-sized groups based on their clinical characteristics or personal preferences: 25