背景：肥胖已成为全球亟需解决的重大公共卫生问题。肥胖人群减重后出现的体重循环现象即体重反复反弹则进一步增加了长期代谢负担和慢性疾病的风险，是影响减重效果和临床结局的重要因素，亦是医学减重中迫切需要解决的关键环节。生活方式管理尤其是其中的营养干预是医学减重干预的基础，但现有研究表明即便是执行相同的营养干预减重策略，最终的减重效果依然存在极大的个体差异，且减重后出现体重循环的比例较高。因此深入了解体重循环形成的机制，实现预测体重循环肥胖者的减重效果，有助于制定个性化营养干预策略，提高长期减重效率预防体重循环，并为今后肥胖治疗的临床决策提供理论基础和新思路。

方法：第一部分 获取人群脂肪组织的转录组数据集和小鼠脂肪组织免疫细胞的单细胞转录组数据，通过生信分析和机器学习，关注代谢性炎症的持续和脂肪组织免疫微环境的变化以探究体重循环的成因。建立体重循环小鼠模型，比较高蛋白饮食（high protein diet，HPD）和限能量饮食（calorie restricted diet，CRD）在代谢性炎症中的干预效果及其对体重循环的影响和潜在的分子机制。第二部分 纳入北京协和医院减重门诊体重循环肥胖者180名，受试者接受“3个月HPD + 6个月CRD”的序贯干预。在干预前、中（HPD干预后）、后测量受试者的体成分、基础代谢、血压、腰围和臀围，同时检测血清肝肾脏功能、血脂、炎症因子及激素水平。在干预前、HPD干预结束后，评估心理困扰和饮食行为，使用结构方程模型探索心理困扰、饮食行为和减重效果的关系。第三部分 在基线、HPD-CRD序贯干预中和干预后收集受试者的外周静脉血，对基线期采集的单个核细胞采用普通转录组测序，对干预前、中、后采集的血清采用非靶向代谢组和靶向蛋白组测序，使用时序分析和多组学关联分析，识别对序贯干预响应度预测的标志物和变化规律。

结果：第一部分 （1）高脂饮食和普通饮食交替法成功建立体重循环小鼠模型。在预实验、正式实验和人群皮下脂肪组织的数据集分析中发现，减重效果不佳与局部的代谢性炎症显著相关。（2）经三种机器学习方法筛选识别了预测体重反弹的关键基因干扰素γ受体1（interferon gamma receptor 1，IFNGR1）和Secernin 2（SCRN2）。（3）对小鼠附睾脂肪组织单细胞数据集分析及结合分子生物学实验发现，巨噬细胞的富集丰度较高，说明巨噬细胞是诱导代谢性炎症的关键细胞，并发现IFNGR1^high自然杀伤细胞（natural killer cell，NK细胞）对代谢性炎症和体重反弹可能存在抑制作用。（4）HPD或CRD均能一定程度的减轻全身和脂肪组织局部Toll 样受体 4（Toll-like receptor 4，TLR4）- NOD样受体热蛋白结构域相关蛋白3（NOD-like receptor thermal protein domain associated protein 3, NLRP3）通路相关的代谢性炎症，与体重循环组小鼠相比，HPD或CRD干预后，白色脂肪中NLRP3、TLR4的蛋白水平显著降低（P＜0.05）。HPD和CRD均能减轻Wnt（Wingless / Integrated）信号通路的激活。 第二部分 共153例受试者完成了该研究的HPD-CRD序贯干预。（1）HPD后，体重循环肥胖者的体重、体脂肪占比、腰臀比、内脏脂肪指数显著降低，肌肉量占比、血糖、血脂、血压也得到显著改善（P＜0.05）。（2）HPD-CRD后，监测指标出现了一定程度的反弹，含体重、体脂肪、骨骼肌指数、血压、血脂，但与基线期相比仍旧呈现明显的改善（P＜0.05）。（3）HPD干预期间，心理困扰与问题饮食行为有关（P＜0.05），饮食行为改善预示着更大的体重减轻（P＜0.05）。间接效应（心理困扰®饮食行为®体重减轻）有统计学意义（P＜0.05）。第三部分 （1）与基线相比，受试者HPD-CRD干预后血清代谢物和脂质发生明显变化。与基线相比，HPD后，主要有不饱和脂肪酸类、磷脂类、氨基酸类、肉碱类和部分抗氧化代谢产物上调和三酰甘油类、溶血磷脂酰胆碱类、磷脂酰肌醇类、鞘磷脂类、胆固醇酯类下调；HPD-CRD干预后后，显著上调了磷脂酰乙醇胺类、磷脂酰胆碱类、磷脂酰肌醇类、鞘磷脂醇胺和下调了三酰甘油类、二酰甘油类、磷脂酰肌醇类。（2）使用聚类的时序分析进一步检查了这期间显著变化的代谢物，确定了四个主要的纵向轨迹簇，并描绘响应这一序贯干预的不同模式。（3）受试者的基线血清炎症蛋白组主要以经典炎症通路为主；HPD后，还涉及了C型凝集素和T细胞受体的信号通路，说明在HPD干预下提升了免疫系统抗感染水平；这一作用在序贯干预结束后仍有体现。（4）根据筛选后差异基因，将数据分为训练集和测试集，EEF1E1和CALM2被确定为与HPD-CRD序贯干预响应度结局的预测标记物。（5）多组学相关网络中，代谢脂质组涉及27个代谢物和脂质，以磷脂酰胆碱和溶血磷脂酰胆碱为主，此外包含单不饱和脂肪酸、多不饱和脂肪酸、肉碱和色氨酸代谢物。炎症蛋白组当中有38个蛋白被该网络纳入，以细胞趋化因子和白介素家族为主。转录组中纳入158个基因，涉及有关糖脂代谢稳态、脂肪生成和代谢性炎症的通路。

结论: （1）体重循环形成与经典型巨噬细胞（classical activated macrophages，M1）和NK细胞长期炎性浸润而引发的代谢性炎症密切相关，IFNGR1可能是脂肪组织中预测长期体重控制结局的标志物。（2）体重循环肥胖者心理困扰与体重控制水平有关，饮食行为可能调节了心理健康和体重控制之间的关系。（3）HPD-CRD序贯干预能够有效的减轻体重循环肥胖者的代谢性炎症，减轻体重、体脂率，增加肌肉含量，提高血糖调节水平，改善血脂和血压。这可能与降低TLR4-NLRP3信号通路和激活Wnt通路有关。（4）EEF1E1和CALM2可能预测体重循环肥胖者对HPD-CRD序贯干预的响应度。其响应的差异性，与基线期和干预后较低的代谢性炎症落差，以及较弱蛋白质/氨基酸代谢敏感度有关。

Background: Obesity has become a significant global public health issue that must be addressed immediately. The weight cycle phenomenon that occurs in obese people after weight loss, that is repeated weight rebound, further increases the long-term metabolic burden and the risk of chronic diseases. It is an important factor affecting the weight loss effect and clinical outcome, and also a key link that needs to be urgently solved in medical weight loss. Lifestyle management, especially nutrition intervention, is the basis of medical weight loss intervention. However, existing studies have shown that even if the same nutrition intervention weight loss strategy is implemented, the final weight loss effect still has great individual differences, and the proportion of weight cycling after weight loss is high. Therefore, an in-depth understanding of the mechanism of the formation of weight cycling and the realization of predicting the weight loss effect of obese people with weight cycling are helpful to develop personalized nutrition intervention strategies, improve the efficiency of long-term weight loss to prevent weight cycling, and provide a theoretical basis and new ideas for future clinical decision-making of obesity treatment.

Methods: The first phase involved collecting transcriptomic datasets from human adipose tissue and single-cell transcriptomic data from immune cells in mouse adipose tissue. The study investigated the causes of weight cycling using bioinformatics analysis and machine learning, focusing on metaflammation persistence and adipose tissue immune microenvironment changes. Mice were established a weight cycling model used to compare the effects of high-protein diet (HPD) and calorie restriction diet (CRD) on metaflammation in weight cycling, as well as the underlying molecular mechanisms. The second phase involved enrolled 180 participants with weight cycling obesity from Peking Union Medical College Hospital. The participants followed a three-month HPD and a six-month CRD. Body composition, basal metabolism, blood pressure, waist and hip circumferences were measured before, during (after HPD intervention), and after the HPD-CRD intervention. Venous blood was drawn for liver and kidney function testing, lipid profiles, inflammatory markers, and hormone levels. Moreover, psychological distress and eating behaviors were measured before and after the high-protein diet phase, with structural equation modeling used to investigate relationships among psychological distress, eating behaviors, and weight loss outcomes. Finally, the third phase obtained peripheral venous blood from the participants using bulk transcriptomics sequencing for single-nucleated cells collected at the baseline and untargeted metabolomics and targeted proteomics sequencing for serum collected before, during, and after sequential intervention. Time-series and multi-omics correlation analyses were used to investigate biomarkers and patterns that predict responsiveness to sequential interventions.

Results: The following were the results in the first phase: (1) A mouse weight-cycling model was successfully established by alternating between high-fat and regular diets. A comparison among the pre-experimental, formal experimental, and human subcutaneous adipose tissue datasets determined that poor weight loss outcomes were significantly associated with localized metaflammation. (2) Three machine learning methods were used to identify key genes that predict weight rebound: interferon gamma receptor 1 (IFNGR1) and Secernin 2 (SCRN2). (3) The analysis of single-cell datasets from mouse epididymal adipose tissue and molecular biology experiments discovered that macrophages as key cells inducing metaflammation and suggested a possible suppressive role of IFNGR1^highnatural killer cells (NK cells) on metaflammation and weight rebound. (4) HPD or CRD interventions reduced metaflammation in systemic and local adipose tissue through the Toll-like receptor 4(TLR4)-NOD-like receptor thermal protein domain associated protein 3(NLRP3) pathway, with significant reductions in NLRP3 and TLR4 protein levels compared to weight cycling mice (P < 0.05). Both interventions reduced Wingless (Wnt) signaling pathway activation. The second phase involved 153 participants who completed the HPD-CRD sequential intervention. The following were the results: (1) The three-month HPD intervention resulted in significant reductions in weight, body fat percentage, waist-to-hip ratio, and visceral fat index, as well as substantial improvements in muscle mass percentage (P < 0.05). Significant improvements were observed in blood glucose, lipid profiles, and blood pressure (P < 0.05). (2) After six months of CRD, weight, body fat, muscle index, blood pressure, heart rate, and lipid profiles showed some rebound but remained significantly improved from baseline (P < 0.05). (3) During the HPD, psychological distress was associated with problematic eating behaviors (P < 0.05), while improvements in eating behavior predicted greater weight loss (P < 0.05). Meanwhile, the indirect effect (psychological distress → eating behavior → weight loss) was statistically significant (P < 0.05). It suggested that eating behavior mediated the relationship between psychological distress and weight loss and that psychological distress mediated weight loss through eating behavior. The following were the results in the third phase: (1) Significant changes in serum metabolites and lipids were observed compared to baseline after HPD-CRD intervention. After HPD, the upregulation of unsaturated fatty acids, phospholipids, amino acids, carnitines, and some antioxidative metabolites and the downregulation of triglycerides, lysophosphatidylcholines, phosphatidylinositols, sphingomyelins, and cholesterol esters. Following HPD-CRD, phosphatidylethanolamines, phosphatidylcholines, phosphatidylinositols, and sphingomyelin levels upregulated significantly, while triglycerides, diglycerides, and phosphatidylinositols levels downregulated. (2) clustered time-series analysis further examined these significantly changed metabolites and identified four main longitudinal trajectory clusters, depicting different response patterns to the sequential intervention. (3) After three months of high-protein and HPD-CRD sequential intervention, classical pathways dominated the baseline serum inflammatory proteome. After the HPD, C-type lectin, and T-cell receptor signaling pathways were involved, indicating an enhanced anti-infection level of the immune system under nutritional intervention with HPD. This immune protection remained evident after completing the HPD-CRD. (4) Data were divided into training and test sets based on the screened differential genes and identified EEF1E1 and CALM2 as predictive biomarkers for responsiveness to HPD-CRD sequential intervention. (5) The metabolic lipidome in the multi-omics correlation network included 27 metabolites and lipids, primarily phosphatidylcholines and lysophosphatidylcholines, including monounsaturated fatty acids, polyunsaturated fatty acids, carnitines, and tryptophan metabolites. The inflammatory proteome consisted of 38 proteins, primarily chemokines and the interleukin family. The transcriptome contained 158 genes involved in glucose and lipid metabolic homeostasis, lipogenesis, and chronic inflammation.

Conclusions: (1) The formation of weight cycles is closely related to metaflammation caused by long-term inflammatory infiltration of classical activated macrophages (M1) and NK cells. IFNGR1 may be a marker in adipose tissue to predict the outcome of long-term weight control.  (2) The psychological distress of people with weight cycling and obesity is related to the level of weight control. Eating behavior may mediate the relationship between mental health and weight control. (3) HPD-CRD sequential intervention can effectively reduce metaflammation, reduce body weight and body fat rate, increase muscle content, improve blood glucose regulation levels, and improve blood lipids and blood pressure in people with weight cycling and obesity. This may be related to reducing the TLR4-NLRP3 signaling pathway and activating the Wnt pathway. (4) EEF1E1 and CALM2 may predict the response of individuals with weight cycling and obesity to HPD-CRD sequential intervention. The difference in response is related to the lower metaflammation gap between baseline and post-intervention, as well as weaker protein/amino acid metabolic sensitivity.