The Dietary Protein Source Modulates Obesity Development and Induces Distinct Plasma Lipid Profiles and Transcriptome Profiles in Liver and Adipose Tissue Authors: Lene S. Myrmel, Even Fjære, Annette Bernhard, Josef D. Rasinger, Ole Jakob Nøstbakken, Helene Søiland, Philipp Rausch, Martin von Bergen, Ulrike Rolle‐Kampzcyk, Liang Xiao, Hanlin Zhou, Guibo Li, Karsten Kristiansen, Lise Madsen Published: 2026 ABSTRACT ABSTRACTDiets based on dairy or seafood proteins are reported to be less obesogenic than diets based on terrestrial meat proteins. The metabolic pathways underlying these differences, including processes related to thermogenesis, remain incompletely understood. Here, we investigated responses at levels of the transcriptome and the metabolome associated with potentially obesogenic high‐fat, high‐protein diets using casein, cod, or chicken as protein sources in mice kept at thermoneutral conditions or room temperature to evaluate metabolic responses to different housing temperatures. By performing plasma metabolomics analyses combined with RNA sequencing and gene set enrichment analyses of liver, muscle, white‐ and brown adipose tissue, we aimed to decipher the metabolic output and pathways modulated by intake of different dietary proteins. Dietary protein intake of casein, cod, or chicken led to differences in obesity development independent of housing temperature. The diet‐specific phenotype was associated with a distinct plasma lipid profile and transcriptomic profile in liver and adipose tissue. One‐carbon metabolism emerged as a potentially affected key pathway, and the mechanistic target of rapamycin complex 1/2 appeared as a predicted upstream regulator and accordingly suggested to be associated with protein source‐dependent differences in obesity development and related metabolic disturbances. Introduction The global increase in obesity prevalence over the last 50 years is closely linked to changes in food systems and more sedentary behavior [1]. Changes in dietary patterns have been recommended to reduce obesity and related diseases [2, 3, 4]. Several different dietary approaches to prevent or counteract weight gain exist [5]. Among those, high‐protein diets are popular as high dietary protein intake is claimed to suppress weight loss‐induced sensation of hunger [6]. Although still controversial, results from several clinical trials suggest a beneficial effect of increasing the dietary protein content [7, 8], and similar results have been reported in rodent trials [9]. However, observational studies report less consistent effects of high‐protein intake [10, 11]. Nevertheless, the notion that high‐fat diet‐induced obesity is prevented in rodents, or at least strongly attenuated when the dietary protein content is high, is largely based on studies using casein or whey as the protein source [9]. In mice, we have demonstrated that casein is the protein source that most efficiently prevents adiposity when obesity prone C57BL/6J mice are fed high‐fat, high‐protein diets using casein, soy, or filets of cod, beef, chicken, or pork as protein sources at thermoneutral conditions [12]. In humans, epidemiological studies indicate that intake of dairy and vegetarian protein sources protects against obesity, whereas a high intake of meat is linked to weight gain [2, 13, 14]. In accordance with data from human studies, rodent trials demonstrate that diets based on plant, dairy or seafood proteins are less obesogenic compared to diets based on terrestrial meat proteins [12, 15, 16, 17, 18]. The mechanisms underlying these substantial differences in obesity development between different protein sources are not fully understood. The obesity‐protective effect of casein intake in rodents has been found alongside maintenance of a classical brown adipose tissue morphology, when compared to intake of terrestrial meat protein [12]. Mild cold exposure, or housing at ambient temperature (20°C–22°C), has been found to increase energy expenditure and attenuate diet‐induced obesity [19, 20, 21]. Mice with reduced thermogenic capacity due to a dysfunctional brown adipose tissue (BAT) exhibit an obese phenotype at thermoneutrality [22, 23, 24], aggravated by high‐fat feeding [25], but remain lean at room temperature [26]. These findings emphasize the potential impact of ambient temperature on diet‐induced obesity development. However, to what extent ambient temperature can modulate the impact of different protein sources on obesity development is not known. The volume and activity of BAT have been positively associated with improved glucose regulation [27], insulin sensitivity, and lower T2D risk in humans [28], emphasizing the relevance and impact for potential modulation by dietary factors. Multi‐omics signatures reflecting responses to different dietary protein sources could improve the resolution of the phenotype [29, 30, 31, 32]. Targeting transcriptional changes alongside a downstream metabolic signature may disclose mechanistic features linking diet to phenotype. The inclusion of several relevant metabolic tissues, such as liver, adipose tissue and muscle, can also expose tissue‐specific metabolic responses. In this study, we have investigated responses at the level of the transcriptome and metabolome associated with high‐fat, high‐protein diets using casein, cod, or chicken as protein sources in mice kept at thermoneutral conditions or room temperature to evaluate metabolic responses to different housing temperatures. We aimed to decipher the metabolic output and pathways modulated by intake of different dietary protein sources by performing plasma metabolomics analyses combined with RNA sequencing and gene set enrichment analyses of liver, muscle, white‐, and brown adipose tissue. We included an energy‐restricted group fed chicken to limit obesity development to further reveal effects induced solely by differences in obesity and/or energy restriction independent of dietary composition and protein source. Materials and Methods Animal Experiment Eighty male C57BL/6J BomTac mice were received from Taconic (Ejby, Denmark) and acclimated for 5 days at room temperature (22°C±1°C) on a low‐fat reference diet (AIN‐93G, Growth Purified Diet, TestDiet) and water ad libitum. After acclimatization, all mice underwent a noninvasive body scan (Bruker Minispec LF50 Body Composition Analyzer mq 7.5, Bruker Optik GmbH, Ettlingen, Germany) and were distributed into eight experimental groups (n = 10) with similar mean body composition. All mice were single‐housed in individually ventilated cages with 50%±5% relative humidity and a 12:12 h light–dark cycle. Four experimental groups were housed at thermoneutral conditions (30°C±1°C), and the other four groups were housed at room temperature (22°C±1°C). The four experimental groups housed at either thermoneutral conditions or room temperature were given a high‐fat, high‐protein (HFHP) diet with either casein, cod, or chicken as the protein source, and the last group was given an HFHP diet based on chicken, but energy‐restricted to keep body mass similar to the group fed cod as a protein source (Table 1). All animals were fed their respective diets three times a week for 12 weeks, and fresh water was provided once a week. All feed leftovers were recorded, and body weight was measured once per week. Feed efficiency was calculated as body mass gain per total energy intake. After 12 weeks, the experiment was terminated, and all mice were anesthetized using 4% isoflurane (Isoba vet, Schering‐Plough, Denmark) and a Univentor 400 Anesthesia Unit Apparatus, and blood was collected by cardiac puncture into EDTA tubes. Blood was centrifuged at 2500 g for 5 min at 4°C and plasma was separated and stored at −80°C for further metabolomics analyses. Liver, kidneys, muscle (musculus tibialis), epididymal white adipose tissue (eWAT), retroperitoneal WAT (rWAT), inguinal WAT (iWAT), and interscapular brown adipose tissue (iBAT) were dissected, weighed, freeze‐clamped in liquid nitrogen, and stored at −80°C for further analyses. TABLE 1The experimental set‐up.GroupDietHousing temperatureCasein 30°CCaseinThermoneutral (30°C±1°C)Cod 30°CCodChicken 30°CChickenChicken R* 30°CChicken R*Casein 22°CCaseinRoom temperature (22°C±1°C)Cod 22°CCodChicken 22°CChickenChicken R* 22°CChicken R*R* designates restricted access to energy.John Wiley & Sons, Ltd. The experimental set‐up. R* designates restricted access to energy. Ethical Statement The animal experiment was approved by the Norwegian Food Safety Authority (FOTS ID 5358). No adverse effects were observed during the experiment. All animal handling and procedures were performed in accordance with the national and international guidelines for the use of animals in research (Regulation on the use of animals in research, Ministry of Agriculture and Food, first July 2015; according to Directive 2010/63/EU of the European Parliament and of the Council of 22nd September 2010). Animal Diets The experimental diets were all high‐fat, high‐protein diets based on either casein powder (C8654 SIGMA, Merck, Darmstadt, Germany), frozen cod‐, or fresh chicken fillets (Table 2). The cod and chicken fillets were heat‐treated to reach a core temperature of 75°C, freeze‐dried, minced, and pulverized. The pulverized protein sources were added to the diets to obtain a concentration of 400 g crude protein/kg diet. Protein content was determined in the powdered protein sources and feed based on the nitrogen levels measured by the Dumas method using a Leco FP 628 nitrogen analyser (Leco Corporation Svenska AB, Sweden) and a nitrogen‐to‐protein conversion factor of N*5.6 for cod and chicken and N*6.15 for casein [33]. The total fat in the added protein sources was measured gravimetrically after extraction with organic solvents as described earlier [15] and used to adjust the level of corn oil added to each diet to achieve equal amounts of fat in the diet. TABLE 2Dietary composition and analyzed levels of energy, fat and protein in experimental diets. Ingredients (g/kg) CaseinCodChickenCasein472——Cod—482—Chicken——501Sucrose140140140Dextrin40.036.048.2Cellulose50.050.050.0Corn oil248242210L‐Cysteine333t‐butylhydroquinone0.0150.0150.015AIN93G mineral mix353535AIN93VX vitamin mix101010Choline bitartrate2.52.52.5Total100010001000Calculated fat from each protein source a 2.28.440Calculated protein from each protein source a 400400400 Analyzed levels in the diet Energy (kcal/100 g)580580580Fat (g/100 g)252524Protein (g/100 g)40.740.940.4 a The respective protein source used in each diet, either casein, cod, or chicken.John Wiley & Sons, Ltd. Dietary composition and analyzed levels of energy, fat and protein in experimental diets. The respective protein source used in each diet, either casein, cod, or chicken. Body Composition and Fatty Acid Measurements in Liver Whole body composition, including total fat mass, lean mass, and free water, was measured by a non‐invasive scanning using a time‐domain nuclear magnetic resonance system (Bruker Minispec LF50 Body Composition Analyzer mq 7.5, Bruker Optic GmbH, Ettlingen, Germany) as described earlier [34]. This was performed for all mice before the start of the experiment and in week 9 of the experimental feeding period. The composition of fatty acids in liver was measured in all mice at termination using gas chromatography coupled with a flame ionization detector as earlier described [18, 35]. Glucose‐ and Insulin Tolerance Tests An oral glucose tolerance test (OGTT) and an insulin tolerance test (ITT) were performed in all mice in week 9 and 10 of the experimental feeding, respectively. The OGTT was performed in mice fasted for 5 h, and 3 mg glucose/g lean body mass was given by oral gavage. ITT was performed in fed mice given 1 U insulin/kg lean body mass by an intraperitoneal injection using short‐acting insulin (Actrapid). For both OGTT and ITT, glucose levels were measured in full blood collected from the tail vein at the indicated time points using a glucometer (Ascensia Contour, Bayer, Norway). During OGTT, additional blood (20 µL) was sampled from the tail vein at time points 0, 15, and 30 min after the oral glucose dose was given to measure plasma insulin using an ELISA kit (Ultra Sensitive Mouse Insulin ELISA Kit, Crystal Chem). Histology and Immunohistochemistry Part of the interscapular brown adipose tissue (iBAT) dissected out at termination was fixed overnight at 4°C in 4% formaldehyde dissolved in 0.1 M phosphate buffer, further dehydrated, cleared in xylene, embedded in paraffin and sectioned (5 µm thick) as described earlier [36]. The immunohistochemical detection of uncoupling protein 1 (UCP1) was performed as described earlier [37], and quantification of the UCP1‐stained areas in the micrographs was performed using ImageJ, Fiji [38]. Plasma Metabolomics Metabolome analyses were carried out with the AbsoluteIDQ p180 Kit (Biocrates Life Science AG, Austria). The Kit identifies and quantifies 188 metabolites from five compound classes (acyl carnitines, amino acids, glycerophospho‐ and sphingolipids, biogenic amines, and hexoses). Amino acids and biogenic amines were analyzed by liquid chromatography (LC)–mass spectrometry (MS) (LC–MS/MS). The substance classes of acylcarnitines, phosphatidylcholines (PCs) (including lysophosphatidylcholines), sphingomyelines, and hexoses were analyzed by flow injection analysis (FIA)‐MS/MS measurements (FIA‐MS/MS). Plasma samples (10 µL) stored at ‐80°C were mixed with an isotopically labeled internal standard in a multititer plate and dried under nitrogen (nitrogen evaporator 96‐well plate, VLM GmbH, Bielefeld, Germany). Afterwards, the metabolites were derivatized with phenylisothiocyanate (PITC) 5% for 20 min at room temperature and subsequently dried for 30 min under nitrogen flow. For extraction, first 300 µL of extraction solvent (5 mM ammonium acetate in methanol) were added and incubated with shaking at 450 rpm (Thermomixer comfort Eppendorf, Hamburg, Germany) for 30 min at room temperature, followed by filtration by centrifugation (Sigma, Taufkirchen) for 2 min at 5000 g. Subsequently, 200 µL were removed from the filtrate, transferred to a fresh multititer deep‐well plate and diluted with 200 µL of water for LC–MS analysis of biogenic amines and amino acids. To the remaining 100 µL from the filtrate, 500 µL of MS running solvent was added for FIA‐MS/MS. Both types of measurements were performed on a QTRAP MS applying electrospray ionization (ESI) (ABI Sciex API5500Q‐TRAP). The MS was coupled to an ultra‐performance liquid chromatography (UPLC) (Waters Acquity, Waters Corporation, Milford, USA). In case of LC–MS, the metabolites were separated by a hyphenated reverse phase column (Agilent, Zorbax Eclipse XDB C18, 3.0 9 100 mm, 3.5 µm; Agilent Waldbronn, Germany) preceded with a precolumn (Security Guard, Phenomenex, C18, 4 9 3 mm; Phenomenex, Aschaffenburg, Germany) applying a gradient of solvent A (formic acid 0.2% in water) and solvent B (formic acid 0.2% in acetonitrile) over 7.3 min (0.5 min 0% B, 5 min 70% B, 0.3 min 70% B, 2 min 0% B) at a flow rate of 500 µL/min. Oven temperature was 50°C. For LC– MS analysis, 10 µL and for FIA, 20 µL were subjected to measurements in the positive and negative modes, respectively. Identification and quantification were achieved by multiple reaction monitoring (MRM) standardized by applying spiked‐in isotopically labeled standards in the positive and negative modes, respectively. For calibration, a calibrator mix consisting of seven different concentrations was used. Quality controls deriving from lyophilized human plasma samples were included for three different concentration levels. For FIA an isocratic method was used (100% organic running solvent) with varying flow conditions (0 min, 30 µL/min; 1.6 min 30 µL/min; 2.4 min, 200 µL/min; 2.8 min, 200 µL/min; 3 min 30 µL/min), and the MS settings were as follows: scan time 0.5 s, IS voltage for positive mode 5500 V, for negative mode −4500 V, source temperature 200°C, nitrogen as collision gas medium; the corresponding parameters for LC–MS were: scan time 0.5 s, source temperature 500°C, nitrogen as collision gas medium. All reagents used in the processing and analysis were of LC–MS grade, unless otherwise stated. Milli‐Q Water ultrapure was used fresh after being prepared by the high‐purity water system by Merck KGaA (Darmstadt, Germany). LC–MS grade acetonitrile (1.00029.2500) and methanol (1.000971.500), as well as pyridine for analysis (1.09728.0100) and formic acid (98%–100%; 1.000263.1000) were purchased from Merck KGaA. PITC (P10034‐10) and ammonium acetate (81.7838‐50) were purchased from Sigma–Aldrich Chemie GmbH (Steinheim, Germany). For the LC‐MS assay, the data analysis and calculation of the concentrations were performed in the Analyst Software (1.6.2). The obtained LC results were then imported as .rdb file into MetIDQ software and assessed together with results from the FIA assay. The software is an integrated part of the kit. This streamlines data analysis by automated calculation of metabolite concentrations providing quality measures and quantification. RNA Isolation RNA was isolated from liver, iWAT, iBAT, and muscle tissue and snap frozen at ‐80°C. RNA isolation was performed using the Agencourt RNAdvance Tissue Kit (Beckman Coulter, USA) and a Biomek NXp robot (Beckman Coulter, USA). Tissue (20‐50 mg) was homogenized in Precellys tubes with beads and 600 µL of lysis buffer using a Precellys 24 lysis & homogenization apparatus (Bertin Technologies, France) at 6000 rpm for 10 s, three times. The tubes were centrifuged at 12 000 g at 4°C for 5 min, and the clear lysate was transferred to an Eppendorf tube and then frozen at −80°C. Before the RNA purification process was carried out on the robot, the samples were thawed and centrifuged a second time at 12 000 g at 4°C for 5 min. 400 µL of the clear lysate was added onto a 96‐well plate suited for the robot, and a purification program was carried out by the robot. Using the robot, 400 µL of binding buffer was added to each lysate, incubated at room temperature for 5 min, kept on a magnet tray for 6 min, and then the supernatant was removed, and 800 µL of wash buffer was added. The samples in the robot were moved to the magnet tray for 5 min, and then the supernatant was removed. 600 µL of 70% ethanol was used to wash twice, and the samples were incubated at room temperature for 10 min. To elute the RNA, 40 µL of dH20 were added to each sample, followed by incubation for 2 min. Finally, samples were placed on the magnet tray for 7 min, and RNA was transferred to a new 96‐well plate. The RNA concentrations (A260/A280 and A260/A230 ratios) were measured using a Nanodrop ND‐1000 spectrophotometer (Saveen Werner, Sweden). An Agilent 2100 Bioanalyzer (Agilent Technologies, USA) instrument was used together with an RNA 6000 Nano LabChip kit (Agilent Technologies, Germany) to validate the quality of the isolated RNA following the protocol provided by the manufacturer. RNA Sequencing Isolated RNA was sequenced at the sequencing company BGI‐Shenzhen Co., Ltd. using the DNBSEQ sequencing platform and paired‐end sequencing [39]. Analyses of Differentially Expressed Genes and Enrichment Analysis In this study, the RSEM software was used for gene expression quantification [40]. We determined the differentially expressed genes (DEGs) between tissues from the different diet groups using DEseq2 [41]. The screening threshold was |logFC|>1, Padj < 0.05. The R package “ClusterProfile” was used to process the Gene Ontology (GO) enrichment analysis for the DEGs. Gene Set Enrichment Analyses (GSEA) and Ingenuity Pathway Analysis (IPA) To determine which biological pathways were affected between the different diets (ANOVA, p < 0.05) and to identify pathways correlating with fat mass, gene set enrichment analyses (GSEA) were performed using the Qlucore omics explorer (Qlucore AB, Version 3.9.11), and pathway analyses were performed using the Kyoto Encyclopedia of Genes and Genomes (KEGG; version 98.0, released on April 1, 2021) functional pathways. Pathways with a normalized enrichment score (NES) > 1.3 (positive or negative values) from the gene set enrichment analyses (GSEA) were considered. The software Ingenuity Pathway Analysis (IPA, Qiagen, Redwood City, CA, USA) was used to perform network analysis with settings according to the species mouse, the relevant tissue for analysis and the rest set to default settings, applying its “core analysis” and further “causal networks” to identify upstream factors that affect expression of the genes in the RNA seq data from liver, iWAT and iBAT. In each comparison, two groups were compared: casein versus chicken, casein versus cod, and chicken versus cod, providing t‐test results from Qlucore that were uploaded to IPA with corresponding log fold change. The false discovery rate (FDR) adjusted p‐values (q < 0.05) were used. Statistical Analysis Physiological data are presented as mean ± SEM. Two‐way ANOVA using diet and temperature as factors with Tukey's multiple comparisons as a post hoc test was performed using GraphPad Prism version 10.0.0 (GraphPad software, La Jolla, California, USA). Outliers were identified using Grubbs test. D'Agostino‐Pearson test was used to test normality, and the Brown–Forsythe test was used to test equality of group variances. Statistical results from the two‐way ANOVA are demonstrated in each graph, and statistical differences (p < 0.05) between the diets (casein, cod, chicken, chicken‐restricted) are marked by different letters. Statistical differences between housing temperatures within the same protein source are denoted with an asterisk. Multivariate statistical analyses, ranked regression, and the plotting of analysis outputs were performed using the Qlucore omics explorer (2023 Qlucore AB, Version 3.9.11). In short, metabolite and gene expression data were subjected to principal component analyses (PCA). Metabolite and transcriptomics data with significantly altered abundance profiles between groups (ANOVA; q < 0.05) were subjected to hierarchical clustering analysis, and in the statistical analyses of these data, we have used a false discovery rate (FDR) adjusted p‐value (q < 0.05). Results Obesity Development Is Determined by Dietary Protein Source Independent of Housing Temperature Feeding mice high‐fat, high‐protein diets based on protein from either casein, cod, or chicken at thermoneutrality led to expected significant differences in body weight development (Figure 1A) in agreement with an earlier study [12]. The mice fed the diet based on chicken gained the most weight during the experimental feeding period and had a higher body weight gain compared to mice fed casein and cod (Figure 1B). Fat mass measured in week nine of the feeding trial reflected the body weight differences (Figure 1C), demonstrating that mice fed casein remained lean, mice fed cod gained more fat mass, whereas chicken‐fed mice gained more fat mass than both casein‐ and cod‐fed mice. The lean mass was higher in mice fed cod and chicken compared to casein‐fed mice (Figure 1D). The housing temperature did not impact any of the parameters for mice fed casein, cod, or chicken ad libitum. Energy restriction prevented the chicken‐induced increase in body weight and fat mass. Body weight and feed intake. (a) Body weight development (n = 10), (b) body weight gain, (c) fat mass, (d) lean mass, (e) energy intake, and (f) feed efficiency after 9 weeks of feeding mice high‐fat, high‐protein diets based on casein, cod, or chicken protein. Statistical results from two‐way ANOVA are provided in the box with each graph; differences between diets are indicated by different letters, whereas statistical differences by housing temperature within each diet are marked with an asterisk (p < 0.05). Two‐way ANOVA of the body weight development curve: Interaction p < 0.0001, Time p < 0.0001, Group p < 0.0001. Chicken R: mice given an energy‐restricted chicken diet. Feed Efficiency Reflects Obesity Development, and Energy Intake is Increased by Thermal Stress All mice housed at room temperature had higher energy intake than mice kept at thermoneutral conditions (Figure 1E), consistent with earlier studies reflecting the impact of mild cold stress on overall energy expenditure [19, 20]. Overall, feed efficiency reflected the differences in obesity development between the diets, demonstrating a marked effect based on the dietary protein source in ad libitum‐fed mice, with the lowest feed efficiency observed in mice fed casein, followed by mice fed cod, and with mice fed chicken having the highest feed efficiency (Figure 1F). White Adipose Tissue Mass Is Increased by Chicken Intake The masses of different white adipose tissue depots after 12 weeks of feeding reflected the pattern in total fat mass and body weight, demonstrating that chicken intake induced an increase in white fat mass (Figure 2A–C). In contrast, the housing temperature did not impact the mass of white adipose tissue depots. As a measure of brown adipose tissue activation, the classical interscapular brown adipose tissue (iBAT) was dissected out, and UCP1 expression was quantified. iBAT mass was higher in mice fed chicken and cod compared to casein (Figure 2D). In all mice fed ad libitum, mild thermal stress at room temperature reduced iBAT mass, whereas UCP1 expression measured by immunohistochemistry was higher in mice at room temperature than in mice at thermoneutrality (Figure 2E). As demonstrated earlier [12], UCP1 expression was elevated in mice fed casein compared to chicken. However, the total UCP1 expression relative to iBAT mass demonstrated an opposite trend and was increased in mice fed chicken compared to casein (Figure 2F), while mild cold exposure only increased total UCP1 expression relative to iBAT mass in mice fed chicken. Adipose tissue and liver mass. Mass of (a) inguinal white adipose tissue (iWAT), (b) epididymal WAT (eWAT), (c) retroperitoneal WAT (rWAT), and (d) interscapular brown adipose (iBAT) tissue dissected out at termination after 12 weeks of feeding mice high‐fat, high‐protein diets based on casein, cod, or chicken protein, housed at 22°C or 30°C. (e) % area of UCP1 expression in iBAT tissue sections (n = 6) immunohistochemically stained with a UCP1 antibody, and (f) the UCP1 expression from immunohistochemical staining relative to the total iBAT weight (n = 6). Statistical results from two‐way ANOVA are provided in the box with each graph; differences between diets are indicated by different letters, whereas statistical differences by housing temperature within each diet are marked with an asterisk (p < 0.05). Chicken R: mice given an energy‐restricted chicken diet. Liver Mass and Fat Content Are Increased by Chicken Intake An interaction effect was seen for liver mass, being affected by both housing temperature and diet. Compared with casein, cod and chicken intake led to increased liver masses following the same pattern as white fat mass (Figure 3A). Compared with room temperature, thermoneutral housing led to lower liver mass in mice fed the energy‐restricted chicken diet. The level of fatty acids in the liver was higher in mice fed chicken compared to all the other diet groups (Figure 3B). Thermoneutral housing induced a significant elevation in hepatic fatty acids in mice fed chicken ad libitum, due to an increase in the saturated‐, monounsaturated‐, and n‐6 polyunsaturated fatty acids (Figure 3C–F and Table 3). Housing at low temperatures, calorie restriction, and dietary protein intake based on casein and cod were all factors preventing hepatic fat accumulation in this study. Liver weight and fatty acids in liver. Mass of (a) liver, (b) total fatty acids in liver, (c) saturated‐, (d) monounsaturated‐, (e) n‐6 polyunsaturated, and (f) n‐3 polyunsaturated fatty acids in liver at termination after 12 weeks of feeding mice high‐fat, high‐protein diets based on casein, cod, or chicken protein, housed at 22°C or 30°C. Statistical results from two‐way ANOVA are provided in the box; differences between diets are indicated by different letters, whereas statistical differences by housing temperature within each diet are marked with an asterisk (p < 0.05). Chicken R: mice given an energy‐restricted chicken diet. TABLE 3Fatty acids in liver.CaseinCodChickenChicken R2‐way ANOVAFA (mg/g)22°C30°C22°C30°C22°C30°C22°C30°CInteractionTemp.Diet16:07 ± 27 ± 26 ± 16 ± 19 ± 3220 ± 146 ± 17 ± 20.00020.0022< 0.000118:03.7 ± 0.53.5 ± 0.83.7 ± 0.33.6 ± 0.63.9 ± 0.34.3 ± 0.83.5 ± 0.53.6 ± 0.7nsns0.020418:1 n‐9 (Oleic acid)7 ± 37 ± 25 ± 25 ± 19 ± 325 ± 195 ± 16 ± 30.0010.0114< 0.000118:2 n‐6 (Linoleic acid)11 ± 312 ± 411 ± 212 ± 316 ± 440 ± 2710 ± 212 ± 60.00120.0034< 0.000120:4 n‐6 (ARA)6 ± 16 ± 13.5 ± 0.43.5 ± 0.55.6 ± 0.47 ± 25.0 ± 0.86 ± 1nsns< 0.000118:3 n‐3 (α‐Linolenic acid)0.06 ± 0.020.05 ± 0.030.07 ± 0.020.06 ± 0.030.15 ± 0.050.4 ± 0.40.07 ± 0.030.09 ± 0.070.00420.0274< 0.000120:5 n‐3 (EPA)<0.01<0.010.5 ± 0.10.5 ± 0.10.11 ± 0.050.2 ± 0.20.07 ± 0.020.07 ± 0.040.0392ns< 0.000122:6 n‐3 (DHA)1.6 ± 0.21.6 ± 0.44.0 ± 0.54.1 ± 0.82.4 ± 0.33 ± 12.1 ± 0.42.6 ± 0.7ns0.0101< 0.0001John Wiley & Sons, Ltd. Fatty acids in liver. Both Dietary Protein Source and Housing Temperature Modulate Glucose Tolerance and Insulin Sensitivity Blood glucose levels during the oral glucose tolerance test (OGTT) performed in week 9 of feeding demonstrated an effect of housing temperature in agreement with earlier studies [42, 43] (Figure 4A,B). Fasting blood glucose and plasma insulin levels were elevated in mice fed the chicken diet compared to the other diets (Figure 4C). In chicken‐fed mice, thermoneutral housing elevated plasma insulin levels in the fasted state. An overall increase in plasma insulin levels during the OGTT was evident in groups fed chicken ad libitum compared to all other diet groups. Furthermore, an ITT performed in week 10 of feeding demonstrated reduced insulin sensitivity in mice fed chicken compared to mice fed casein (Figure 4D). The insulin‐induced lowering of glucose levels in blood during the ITT demonstrated an increased response in mice fed casein. Glucose tolerance and insulin sensitivity. (a) Blood glucose levels during the oral glucose tolerance test (OGTT) and (b) the area under the curve in week 9 of feeding mice high‐fat, high‐protein diets based on casein, cod, or chicken protein, housed at 22°C or 30°C. (c) Blood glucose and plasma insulin after 5 h fasting, and plasma insulin after 15 and 30 min during the OGTT. (d) Glucose levels measured during an insulin tolerance test (ITT) in week 10 and the change in blood glucose levels 15 and 30 min after insulin injection. Statistical results from two‐way ANOVA are provided in the box with each graph; differences between diets are indicated by different letters, whereas statistical differences by housing temperature within each diet are marked with an asterisk (p < 0.05). Two‐way ANOVA of the OGTT curve: Interaction ns, Time p < 0.0001, Group p < 0.0001. Two‐way ANOVA of the ITT curve: Interaction ns, Time p < 0.0001, Group p < 0.0001. Chicken R: mice given an energy‐restricted chicken diet. The Plasma Metabolome Profile Is Determined by the Dietary Protein Source The plasma metabolome represents a functional indicator of metabolism [44] and is also linked to disease risk [45]. The lipid‐specific plasma metabolome measured by flow injection analysis‐MS/MS measurements (FIA‐MS/MS) (Table S1) demonstrated a separation between mice given the different diets, independent of housing temperatures (Figure 5A). Plasma metabolomics (FIA platform) profiles. (a) Principal component analysis (PCA) including all 170 identified metabolites from mice fed different high‐fat, high‐protein diets based on casein, cod, or chicken protein, housed at 22°C or 30°C (n = 9–10). (b) Hierarchical clustering analysis (HCA) of plasma metabolites significantly altered in abundance (ANOVA, FDR q < 0.05, 158 metabolites) between the experimental groups. Graphs were made, and statistical analyses were performed using the Qlucore omics explorer. The color scale of the heat map indicates change from the mean in the normalized data (mean = 0, variance = 1). (c) Total phosphatidylcholine (PC) in plasma (µM). Statistical results from two‐way ANOVA are provided in the box; differences between diets are indicated by different letters, whereas statistical differences by housing temperature within each diet are marked with an asterisk (p < 0.05). An overview of all metabolites can be found in Table S1. Chicken R: mice given an energy‐restricted chicken diet. Overall, the abundances of 158 out of 170 identified metabolites/features were differently affected (q < 0.05) between the eight experimental groups (Table S2). Hierarchical clustering analysis of the differently affected features demonstrated a diet‐specific pattern (Figure 5B). The metabolites differing in abundances between the experimental groups comprised acylcarnitines, sphingolipids, and glycerophospholipids, including especially the PCs. Total PCs were higher in plasma from mice fed chicken ad libitum compared to mice fed casein (Figure 5C). Between the groups fed different diets ad libitum, 147 lipid‐specific plasma metabolites/features differed in abundances (Table S3), whereas only 49 variables differed in abundances between mice housed at different temperatures (Table S4). However, plasma metabolome profiles overlapped across housing temperatures within each diet group (Figure 5A,B), indicating that dietary protein composition had a stronger effect than temperature regimen on plasma metabolome composition. Energy restriction led to a profound change in the lipid‐specific metabolite profile, where 105 variables exhibited significantly different abundances between mice fed chicken ad libitum and mice fed chicken with energy restriction (Table S5). Because the chicken ad libitum and chicken restricted groups consumed diets of identical composition, the 105 lipid‐specific plasma variables that differed in abundance between the two groups likely reflect alterations associated with obesity status and/or the metabolic consequences of energy restriction per se. Plasma measurements of amino acids, biogenic amines, and related features on the LC platform identified 33 metabolites with a significantly different abundance between the experimental groups (Figure S1 and Tables S6 and S7). Of these, methionine, glycine, and serine were among the metabolites differing in abundance, and these amino acids are closely linked to metabolites in PC metabolism, which also were found to be significantly affected. Dietary Protein Sources Distinctly Modulate the Transcriptome Profiles of Liver, White, and Brown Adipose Tissue As thermoneutral housing resulted in a more pronounced metabolic response in chicken‐fed mice and better reflects human metabolic responses [46], transcriptomic analyses were restricted to mice housed at thermoneutrality. RNA sequencing of liver, muscle, WAT, and BAT was performed to obtain global transcriptome profiles of tissues relevant for metabolic health in mice fed different high‐fat, high‐protein diets ad libitum (Table S8). To investigate the diet‐induced impact on tissue gene expression, RNA sequencing was performed on mice fed the different dietary protein sources ad libitum. PCA of all RNA transcripts for each tissue demonstrated a marked and global impact of the dietary protein source in liver, iWAT and iBAT (Figure 6A–C). Muscle tissue (m. tibialis) did not exhibit distinct changes in transcriptome profiles in response to the dietary protein source (Figure 6D). RNA seq. Gene expression profiles of (a) liver, (b) iWAT (inguinal white adipose tissue), (c) iBAT (interscapular brown adipose tissue), and (d) muscle (m. tibialis) presented by principal component analysis of all detected RNA transcripts and hierarchical clustering analysis of transcripts significantly altered in abundance between the groups fed different high‐fat, high‐protein diets based on casein, cod, or chicken protein (ANOVA, q < 0.05), housed at 30°C (n = 5). The color scale of the heat map indicates change from the mean in the normalized data (mean = 0, variance = 1). Graphs were made, and statistical analyses were performed using the Qlucore omics explorer. All RNA transcripts in liver, iWAT, iBAT and muscle (m. tibialis) can be found in Table S8. Transcripts significantly altered in abundance (ANOVA, FDR q < 0.05) in liver, iWAT and iBAT are listed in Tables S9–S11. In liver tissue, 441 RNA transcripts were differently regulated (q < 0.05) by the diet. Hierarchical clustering analysis of the differently regulated RNA transcripts demonstrated that especially casein intake led to a distinct transcriptome profile (Figure 6A and Table S9). In iWAT, 2100 RNA transcripts were found to be differently regulated (q < 0.05) between the diets, and mice fed chicken demonstrated the most distinct expression profile (Figure 6B and Table S10). The transcriptome profile in brown adipose tissue of the mice was less affected by the diet, with only 65 differently regulated RNA transcripts being observed (q < 0.05) between the dietary groups, where mice fed chicken had the most distinct transcriptome profile in iBAT (Figure 6C and Table S11). In muscle, no RNA transcripts were found to be significantly modulated by the diets (q < 0.05) (Figure 6D). Gene Set Enrichment Analysis (GSEA) Identifies Several Pathways Differentially Affected Between Mice Fed Different Protein Sources Gene set enrichment analysis (GSEA) of the RNA sequencing data was performed to gain insight into potential pathways affected between the mice given different protein sources. Based on the normalized enrichment score (NES) in liver, the highest‐ranked pathways between mice fed the different diets were amino acid‐ and one‐carbon metabolism; further down the list were pathways involved in lipid metabolism and glucose regulation (Tables 4 and S12, and Figure S2). In iWAT, one‐carbon metabolism and related pathways were also found to rank highest based on the RNA transcripts differing between the diets; the other top pathways listed were within lipid metabolism, especially sphingolipids, glycerolipids, and glycerophospholipids (Table S13 and Figure S3). This corresponded with the distinct lipid‐specific plasma profile data (Figures 5C and S1, and Table S2) comprising sphingolipids and glycerophospholipids, including especially the plasma PCs that significantly differed in abundance between the dietary groups. In iBAT, fatty acid biosynthesis and PPAR signaling pathways were the two top pathways differentially affected between the dietary groups based on the RNA sequencing data (Table S14 and Figure S4). Overall, several plasma metabolites and RNA transcripts involved in one‐carbon metabolism and related pathways significantly differed between the dietary groups in liver and adipose tissue (Figure 7). TABLE 4Top ten pathways from gene set enrichment analyses (GSEA) ranked by normalized enrichment score (NES).Top ten pathways suggested by gene set enrichment analyses (GSEA)NameNES p q Liver Tryptophan metabolism1.810< 0.0010.183Amino acid metabolism1.763< 0.0010.112One‐carbon metabolism and related pathways1.7190.0030.094Omega‐3/Omega‐6 FA synthesis1.6960.0010.078Cholesterol metabolism (includes both Bloch and Kandutsch‐Russell pathways)1.6640.0290.071Complement activation, classical pathway1.6570.0020.061Eicosanoid metabolism via Cytochrome P450 Mono‐Oxygenases (CYP) pathway1.6130.0020.067Glycolysis and gluconeogenesis1.6090.0010.059Metapathway biotransformation1.607< 0.0010.054One‐carbon metabolism1.586< 0.0010.056 Inguinal white adipose tissue (iWAT) One‐carbon metabolism and related pathways1.8700.0020.073Sphingolipid metabolism (integrated pathway)1.824< 0.0010.048Sphingolipid metabolism (general overview)1.820< 0.0010.033Prostaglandin synthesis and regulation1.744< 0.0010.027Selenium micronutrient network1.733< 0.0010.026Fatty acid biosynthesis1.720< 0.0010.026Glycerolipids and glycerophospholipids1.694< 0.0010.027Oxidative damage1.677< 0.0010.028Glutathione and one‐carbon metabolism1.650< 0.0010.033Oxidative stress1.6480.0010.032 Interscapular brown adipose tissue (iBAT) Fatty acid biosynthesis1.8250.0020.069PPAR signaling pathway1.7080.0020.113Glutathione metabolism1.6100.0020.162Cholesterol metabolism (includes both Bloch and Kandutsch‐Russell pathways)1.6090.0020.122Mitochondrial LC‐Fatty acid beta‐oxidation1.6070.0230.099Cytoplasmic ribosomal proteins1.5800.0970.103Cholesterol biosynthesis1.5670.0020.098One‐carbon metabolism1.565< 0.0010.087Folic acid network1.5640.0040.078Glycogen metabolism1.5630.0010.071John Wiley & Sons, Ltd. Top ten pathways from gene set enrichment analyses (GSEA) ranked by normalized enrichment score (NES). Illustration of one‐carbon metabolism and related pathways (https://sandbox.wikipathways.org/pathways/WP1770.html) [47]. Significantly modulated metabolites in plasma and RNA transcripts in liver, inguinal white adipose tissue (iWAT), and interscapular brown adipose tissue (iBAT) between diets are highlighted. Yellow marked with blue line: significantly modulated plasma metabolites (FDR, q < 0.05). Red marked: significantly modulated RNA transcripts in liver (FDR, q < 0.05). Blue marked: significantly modulated RNA transcripts in iWAT (FDR, q < 0.05). Purple marked: significantly modulated RNA transcripts in both liver and iWAT (FDR, q < 0.05). Green marked: significantly modulated RNA transcripts in both iBAT and iWAT (FDR, q < 0.05). Gene Expressions Associated With Dietary Intake and Fat Mass Many of the pathways identified by GSEA as being affected by diet in the liver were also associated with total fat mass, reflecting changed metabolism based on increased fat mass (Figure S5). In iWAT, however, only a subset of the pathways affected by diet correlated with adiposity, suggesting that several of the diet‐affected pathways in this depot were independent of differences in obesity. A similar pattern was observed in iBAT, where many of the pathways affected by fat mass did not correspond to variation in diet. Venn diagrams of all the gene transcripts significantly affected by diet and total fat mass in each tissue demonstrate a great overlap in the gene transcripts affected, but also gene transcripts specifically affected by either diet or fat mass (Figure S5B and Tables S15–S17). In liver, 253 genes were affected by both diet and total fat mass, whereas 188 gene transcripts were specifically affected by diet and 236 by fat mass. In iWAT, overall, more gene transcripts were significantly affected, where 1743 gene transcripts were affected by both diet and fat mass, 357 only by diet, and 988 by fat mass. In iBAT, most of the significantly affected gene transcripts were associated with total fat mass. GSEA indicated that pathways involved in one‐carbon metabolism, amino acid metabolism, glutathione metabolism, tryptophan metabolism, alanine and aspartate metabolism, and oxidative stress in the liver were positively associated with obesity, and genes from these pathways were found to be enriched in mice fed chicken compared to mice fed casein (Figure S6). In iWAT and iBAT, pathways involved in one‐carbon metabolism, glutathione metabolism, folic acid network, and oxidative stress were negatively associated with obesity. Considering glucose regulation and insulin sensitivity, enrichment in gene sets from pathways like glycogen metabolism in iBAT, glycolysis and gluconeogenesis in liver and iWAT, and matrix metalloproteinases in liver were positively associated with obesity (Figure S6). Genes within the leptin–insulin signaling overlap pathway were negatively associated with obesity in liver and white adipose tissue, suggesting a dysregulation in this pathway, which may be linked to the changes demonstrated in insulin sensitivity. Predicted Upstream Regulators Ingenuity Pathway Analysis (IPA) was used to associate the differentially expressed genes between mice fed different protein sources with potential upstream regulators. The differentially expressed genes (FDR q < 0.05) were used as input for “core analysis” in IPA, and a list of regulators was retrieved from the analysis “causal networks,” producing small hierarchical networks of regulators that control the expression of the targets in the dataset and indicating potential upstream regulators. The most distinct gene expression profile was observed in white adipose tissue between mice given casein compared to mice given chicken, indicating several potential upstream regulators (Table S18). The mechanistic target of rapamycin (mTOR)C1/2 complex appeared as a central potential upstream regulator, and other regulators related to this pathway, such as Akt, were also evident within the networks from most suggested regulators. In white adipose tissue from mice fed casein compared to mice fed chicken, the mTORC1/2 complex was predicted to be inhibited (z‐score −4.857), where 132 out of 196 genes had measurement direction consistent with an inhibition (Table S19). A list representing potentially affected upstream regulators between mice given casein and cod in white adipose tissue and liver and between mice fed casein and chicken in liver and iBAT is found in Tables S20–S24. No genes or too few genes with FDR <0.05 were found informative regarding potential upstream regulators between the rest of the comparisons, including between mice fed casein and cod in the white adipose tissue and iBAT and between mice given chicken and cod in liver and iBAT, in addition to all comparisons in muscle. Discussion The present study demonstrates that dietary protein source in a high‐fat, high‐protein diet differentially influences obesity development and glucose tolerance in mice, independent of housing temperature. Mice fed casein, cod, or chicken displayed distinct phenotypic outcomes at both room temperature and thermoneutrality, indicating that the metabolic effects of protein source are not secondary to differences in thermogenic demand. These findings extend previous observations that protein source, rather than protein quantity alone, modulates obesity development despite comparable energy intake [12, 48]. Dietary protein source exerted a pronounced effect on the plasma metabolome, whereas housing temperature had only minor influence. In particular, the abundance of several lipid‐related metabolites differed, including acylcarnitines, sphingolipids, and glycerophospholipids such as PCs. These metabolome differences were paralleled by distinct transcriptome profiles in liver, iWAT, and iBAT, whereas skeletal muscle showed limited responses to dietary protein source. GSEA identified one‐carbon metabolism and related pathways as key regulators associated with obesity development, with opposing regulation in liver versus adipose tissue. In liver, expression of genes involved in one‐carbon metabolism was positively associated with obesity and enriched in mice fed chicken compared to casein. In contrast, in both iWAT and iBAT, this pathway was enriched in casein‐fed mice and negatively associated with obesity. This tissue‐specific divergence suggests fundamentally different roles of one‐carbon metabolism in hepatic versus adipose tissue during diet‐induced obesity, summarized in Figure 8. Experimental factors and outcome. Illustration of the impact of experimental factors, dietary protein source, housing temperature, and energy restriction on affected metabolic parameters, including obesity development, insulin sensitivity, glucose tolerance, hepatic fat accumulation, and UCP expression. A list of the main pathways found by gene set enrichment analysis to be modulated by the diet in liver and adipose tissue, in addition to the main plasma metabolites affected by the diet and the predicted upstream regulator mTOR involved based on pathway analysis. The illustration was created in BioRender (https://BioRender.com). The plasma metabolomic data support these transcriptomic findings. Methionine, glycine, and serine, amino acids directly related to the one‐carbon cycle [49], were reduced in mice fed chicken compared to casein, whereas total plasma PC levels were elevated. These alterations are consistent with previous studies linking high‐fat intake to dysregulated one‐carbon metabolism, hepatic lipid accumulation, insulin resistance, and impaired glucose tolerance [50, 51]. Notably, increased hepatic PC synthesis is closely associated with fatty liver development [52], and diets deficient in methyl donors are known to induce hepatic steatosis [53]. One‐carbon metabolism utilizes dietary choline, serine, and glycine to generate methyl donors, supporting nucleotide synthesis and regulating phospholipid methylation, including PC synthesis [54]. In the present study, the combination of elevated plasma PCs and reduced methionine, glycine, and serine in chicken‐fed mice suggests increased flux through hepatic one‐carbon pathways, potentially contributing to hepatic lipid accumulation and insulin resistance. This interpretation is supported by the increased liver mass, elevated hepatic fatty acids, impaired insulin sensitivity, and the obese phenotype observed in mice fed chicken‐based diets. In addition to lipid metabolism, expression of genes involved in amino acid metabolism, glutathione synthesis, and related pathways was enriched in the livers of chicken‐fed mice compared with casein‐fed mice. One‐carbon metabolism is closely linked to redox homeostasis through the production of glutathione and NADPH [49]. In the present study, expression of genes associated with glutathione metabolism followed the same tissue‐specific pattern as one‐carbon metabolism, being positively associated with obesity in liver but negatively associated in adipose tissue. Similarly, pathways related to oxidative stress showed opposite associations with obesity in liver versus adipose tissue, highlighting divergent metabolic adaptations in response to dietary protein sources in these tissues. Upstream regulator analysis identified the mTORC1/2 complex as a potential mediator of the observed transcriptional differences, particularly in white adipose tissue. mTOR signaling was predicted to be inhibited in mice fed casein compared with chicken, a finding consistent with the anti‐obesogenic phenotype associated with casein intake. mTOR integrates signals from amino acids, energy status, and growth factors to regulate protein synthesis, lipid and nucleotide biosynthesis, mitochondrial function, and autophagy [55, 56, 57, 58]. Chronic hyperactivation of mTORC1 has been implicated in obesity, insulin resistance, and type 2 diabetes [58, 59, 60, 61]. Dietary protein composition is known to influence mTOR activity, particularly through amino acids such as methionine and the branched‐chain amino acids (BCAAs) [56, 62, 63, 64, 65, 66]. Although direct measurements of mTOR activity were not performed, the transcriptome and metabolome profiles observed here strongly suggest differential mTOR signaling in response to protein source intake. In adipose tissue, inhibition of mTOR signaling in casein‐fed mice may contribute to reduced lipid accumulation and improved metabolic outcomes, whereas enhanced mTOR activation in chicken‐fed mice may promote anabolic processes associated with obesity. The potential role of one‐carbon metabolism in modulating mTOR activity further strengthens this interpretation. One‐carbon metabolism supports nucleotide synthesis, amino acid homeostasis, and cellular energy production, processes tightly regulated by mTOR signaling [49, 57]. Altered availability of methyl donors and amino acids in chicken‐fed mice may therefore indirectly enhance mTOR activation, exacerbating obesity‐related metabolic disturbances. It is challenging to disentangle to what extent the observed effects are specific to the obese state, housing temperatures or the dietary protein source. An earlier study in already obese mice demonstrated that caloric restriction reduced obesity irrespective of the protein source, but overall, the dietary protein source had a greater impact in obesity prevention than reversal of obesity [36]. In the present study, energy restriction in mice fed chicken attenuated obesity development and led to improved glucose tolerance and insulin sensitivity compared to the mice fed chicken ad libitum. The plasma metabolome revealed a distinct profile between mice fed chicken ad libitum and mice fed chicken with energy restriction, demonstrating effects induced solely by the differences in obesity and/or energy restriction. Housing temperature influenced several metabolic parameters such as overall glucose tolerance, but some effects were protein source dependent. Thermoneutral housing affected glucose tolerance, but increased fasting insulin secretion only in chicken‐fed mice, consistent with earlier studies demonstrating variable effects of thermoneutral housing on glucose metabolism [42, 43, 67]. Thermoneutrality in combination with chicken intake led to fatty acid accumulation in liver, found to be prevented by housing at room temperature, calorie restriction, or the intake of either casein or cod. This aligns with the liver functioning as a central coordinator of thermogenesis [68] and that ambient temperature could either mask or highlight metabolic phenotypes [46, 69]. These findings highlight the importance of considering both diet composition and housing temperature in metabolic studies. While thermoneutrality is often proposed as a more human‐relevant housing condition [42], the present data demonstrate that its metabolic impact depends strongly on dietary protein context. Notably, casein‐fed mice maintained classical brown adipose tissue morphology and UCP1 expression even at thermoneutrality, consistent with previous findings [12]. However, the differences in iBAT and calculating UCP1 expression per total iBAT mass demonstrated mice fed chicken had the highest UCP1 expression. The specific dietary components responsible for the observed differences between protein sources remain unclear. Variations in amino acid composition, particularly methionine and BCAAs, are likely contributors, given their established roles in mTOR activation and metabolic regulation [62, 64, 65]. Differences in micronutrient content affecting methyl donor availability, including folate, vitamin B12, choline, betaine, and vitamins B2 and B6, may also influence one‐carbon metabolism [67, 70]. Additionally, endogenous fatty acids present in the protein sources, such as n‐3 polyunsaturated fatty acids in phospholipid form, may modulate lipid metabolism and insulin sensitivity [71, 72]. Differences in amino acid digestibility between protein sources could further affect circulating amino acid availability and downstream signaling pathways [73]. Limitations The present study is limited by its associative nature. Although integrated metabolomic and transcriptomic analyses identify one‐carbon metabolism and mTOR signaling as key candidate pathways, causal relationships were not established. Future studies using pharmacological or genetic modulation of these pathways are necessary to determine their direct role in mediating protein source–dependent effects on obesity and glucose tolerance. Additionally, only male mice were studied, and sex‐specific responses to dietary protein sources should be explored [74]. Finally, translation of these findings to humans remains to be determined. Conclusion This study demonstrates that dietary protein source—casein, cod, or chicken—significantly influences obesity development and metabolic health in mice fed a high‐fat, high‐protein diet, independently of housing temperature. Distinct plasma lipid profiles and tissue‐specific transcriptomic responses were associated with these phenotypic differences. One‐carbon metabolism emerged as a central pathway linking protein source to lipid metabolism and obesity, with opposing regulation in liver and adipose tissue. Furthermore, mTORC1/2 signaling was identified as a potential upstream regulator, particularly affected in casein‐fed mice. These findings underscore the importance of protein sources in dietary interventions and highlight metabolic pathways that warrant further mechanistic investigation. Author Contributions LM, KK, LSM, and EF: conceptualization. LSM, EF, and HS: investigation. All authors: data curation and/or formal analysis. LM, LSM, and KK: funding acquisition. LSM and LM: writing – original draft. All authors: writing – review and editing. LSM, LM, EF, HS, AB, JR, KL, URK, MvB, ZX, XL, HZ, GL, OJN, and PR: formal analysis and visualization. LSM: writing – original draft. All authors interpreted the results, edited, revised, and contributed to finalizing the manuscript. Funding This work was funded by the Institute of Marine Research and the Norwegian Ministry of Trade, Industry and Fisheries. Conflicts of Interest The authors declare no conflicting interests, financial or otherwise. The author JD Rasinger declares that he is currently employed with the European Food Safety Authority (EFSA) in the Food Ingredients and Packaging Unit (FIP) that provides scientific and administrative support to the Panel on Food Additives and Flavourings. However, the positions and opinions presented in this article are those of the author alone and do not represent the views/any official position or scientific works of EFSA. The author's main contributions to this article were made before joining EFSA while employed at the Institute of Marine Research. Supporting information Supporting File 1: mnfr70573‐sup‐0001‐SuppFigureS1‐S6.xlsx. Supporting File 2: mnfr70573‐sup‐0002‐SuppTableS1‐S24.xlsx.