Metabolic advances in 2025: from clinical breakthroughs to molecular reprogramming

Abstract

The year 2025 represented a turning point in metabolic research, marked by advances that combined unprecedented clinical efficacy with deep mechanistic insight. Landmark obesity trials redefined therapeutic expectations, with head-to-head and combination studies showing that the depth and distribution of weight loss are critical determinants of metabolic benefit across obesity and type 2 diabetes. In parallel, gene-editing studies crossed a translational threshold, showing that durable modification of metabolic pathways in humans is feasible, from bespoke correction of inborn errors to population-scale lipid lowering.

Mechanistic investigations challenged long-standing assumptions about metabolic regulation. Experimental work revealed that mitochondrial electron transport functions as a dynamic redox regulator rather than a passive energy conduit, linking coenzyme Q imbalance and reverse electron transport to hepatic steatosis and metabolic dysfunction. Other studies reframed nutrient exposure and endogenous metabolites, demonstrating that non-nutritive sweeteners and cyanide exert context-dependent metabolic effects through regulated endocrine and redox pathways.

At the systems level, multi-omics analyses defined reproducible microbiome–metabolome signatures associated with impaired glucose regulation, while artificial intelligence and continuous glucose monitoring exposed dynamic glycemic phenotypes invisible to conventional biomarkers. Precision-nutrition studies further showed that selective manipulation of sulfur amino acid availability can program thermogenic and metabolic responses.

Collectively, these studies illustrate how metabolism in 2025 was approached as a modifiable, programmable system, shaped by clinical intervention, molecular control, and data-driven phenotyping, and point toward an era of increasingly precise and integrated metabolic medicine.

1. Introduction

The year 2025 marked a distinctive phase in metabolic research, defined less by incremental discovery and more by the consolidation of precision across clinical, molecular, and computational domains. A series of landmark studies not only advanced therapeutic efficacy but also reshaped mechanistic understanding, revealing metabolism as a system increasingly amenable to targeted, durable, and programmable intervention.

The studies highlighted in this editorial were selected because each addressed a clearly defined unmet question in metabolic disease, while collectively illustrating a broader shift: metabolic regulation is no longer viewed solely through isolated pathways or single-organ effects, but as an integrated, multi-level process spanning pharmacology, gene editing, nutrient signaling, redox biology and data-driven phenotyping. Rather than attempting a comprehensive survey, this editorial focuses on specific high-impact investigations that exemplify how metabolism advanced in 2025.

Several clinical trials fundamentally redefined expectations in obesity and diabetes, showing that the depth and distribution of weight loss are critical determinants of metabolic benefit. In parallel, gene-editing studies crossed a translational threshold, showing that durable modification of metabolic pathways in humans is feasible. Mechanistic investigations challenged entrenched assumptions regarding mitochondrial function, nutrient toxicity and endogenous metabolic mediators, while multi-omics and artificial intelligence–based approaches refined the interpretation of metabolic heterogeneity at both the individual and disease-specific level.

Taken together, these advances position 2025 as a year in which metabolism was approached not as a static network of pathways, but as a dynamic, programmable system whose behavior can be reshaped with increasing specificity, durability, and clinical relevance.

2. Tirzepatide versus semaglutide: redefining comparative efficacy in obesity pharmacotherapy

The direct comparison of tirzepatide and semaglutide published in 2025 represents a critical advance in obesity pharmacotherapy, not because it introduced a new mechanism, but because it quantified the clinical impact of dual incretin receptor agonism (GIP and GLP-1) relative to single GLP-1 receptor agonism in a head-to-head design, randomized controlled design. [1]. In adults with obesity but without diabetes, tirzepatide (10 mg or 15 mg) administered subcutaneously once weekly for 72 weeks resulted in significantly greater mean percent weight loss (−20.2 %) compared to semaglutide (1.7 mg or 2.4 mg; −13.7 %), with superiority across multiple endpoints including proportions achieving ≥10 %, ≥15 %, ≥20 %, and ≥25 % weight reduction, and greater reduction in waist circumference. The findings are consistent with prior indirect comparisons and real-world cohort studies, which also showed greater weight loss with tirzepatide versus semaglutide [2,3]. Therefore, combined GIP–GLP-1 receptor activation confers incremental benefit beyond GLP-1 receptor agonism alone among individuals with obesity but without diabetes. These effects were accompanied by greater reductions in waist circumference, indicating preferential loss of central adiposity, a parameter closely linked to insulin resistance and cardiometabolic risk. Together, these findings reinforce the principle that the depth and distribution of weight loss are critical determinants of metabolic benefit, particularly with respect to insulin sensitivity, lipid metabolism, and downstream cardiovascular risk.

From a mechanistic perspective, the trial provided indirect but persuasive evidence that GIP receptor engagement modulates energy balance through pathways distinct from appetite suppression alone, potentially involving adipose tissue insulin sensitivity and nutrient partitioning. However, the study was not designed to disentangle these mechanisms, and metabolic phenotyping beyond weight and glycemic indices remained limited. Gastrointestinal adverse events were common in both groups and broadly comparable, suggesting that enhanced efficacy was not offset by disproportionate tolerability concerns.

Taken together, this study establishes tirzepatide as the most effective single-agent pharmacotherapy for obesity reported to date and sets a new benchmark for therapeutic efficacy in the field. At the same time, it highlights unresolved issues regarding long-term durability, cost-effectiveness, and optimal positioning relative to emerging combination regimens, questions that subsequent studies in 2025, including combination approaches such as amylin–incretin co-agonism, began to address.

3. Combination incretin–amylin therapy across obesity and type 2 diabetes: insights from REDEFINE-1 and REDEFINE-2

The REDEFINE-1 and REDEFINE-2 trials published in 2025 collectively established combination amylin–GLP-1 receptor agonism (cagrilintide plus semaglutide, "CagriSema") as a distinct therapeutic advance in metabolic disease by evaluating its efficacy in adults with obesity alone (REDEFINE-1) and with obesity plus type 2 diabetes (REDEFINE-2) [4,5]. Both trials demonstrated that augmenting incretin signaling with an amylin analog provides consistent metabolic benefit across the obesity–diabetes spectrum, with greater weight loss and improved glycemic control compared to semaglutide monotherapy.

In REDEFINE-1 (obesity without diabetes), once-weekly cagrilintide–semaglutide (2.4 mg each) resulted in a mean weight reduction of 20.4 % at 68 weeks, significantly exceeding the effect of semaglutide alone, cagrilintide alone or placebo [4]. In REDEFINE-2 (obesity with type 2 diabetes), the same combination achieved a mean weight reduction of 13.7 % and superior glycemic control, with 73.5 % of patients reaching HbA1c ≤ 6.5 %, compared to 15.9 % with placebo [5]. These results support the mechanistic concept that amylin–incretin co-agonism targets shared pathophysiological mechanisms underlying obesity and diabetes, including complementary effects on appetite regulation and energy balance.

The trials indicate a shift toward combination pharmacotherapy in metabolic disease, as the magnitude of benefit with CagriSema approaches or exceeds that of dual and triple agonists such as tirzepatide and retatrutide [6]. However, both studies raise unresolved questions about long-term tolerability (notably gastrointestinal adverse events and discontinuation rates), durability of effect, and cost–benefit balance compared to other multi-agonist therapies. Further long-term and comparative studies are needed to clarify these issues.

4. Aspartame and atherosclerosis: a mechanistic link through insulin-triggered vascular inflammation

While the previous clinical trials emphasized the therapeutic potential of amplifying hormonal signaling, contemporaneous mechanistic studies underscored how seemingly neutral dietary exposures may exert opposing effects on cardiometabolic risk through endocrine and inflammatory pathways. A notable mechanistic advance in 2025 was the demonstration that a widely used non-nutritive sweetener may aggravate atherosclerosis through a defined neuroendocrine–immune axis. In a study by Wu et al., consumption of 0.15 % aspartame markedly increased insulin secretion in both mice and monkeys, and this effect was abolished by bilateral subdiaphragmatic vagotomy, directly implicating parasympathetic activation in non-glucose-mediated insulin release [7]. In ApoE−/− mice, continuous aspartame exposure exacerbated atherosclerotic plaque formation and growth via an insulin-dependent mechanism, as shown by both gain-of-function (slow-release insulin pumps) and loss-of-function (insulin abrogation) experiments.

The key conceptual step was the identification of the endothelial trigger. Whole-genome profiling of insulin-stimulated arterial endothelial cells identified Cx3cl1 Backspace as the most upregulated gene, linking hyperinsulinemia to endothelial inflammatory signaling. The causal role of this pathway was confirmed by genetic deletion of myeloid Cx3cr1, which completely abrogated aspartame-exacerbated atherosclerosis, establishing the endothelial CX3CL1 chemokine–myeloid CX3CR1 axis as a key inflammatory conduit. The study appropriately notes limitations regarding sweetener specificity and model relevance, but it reframes non-nutritive sweetener exposure as a potential driver of vascular risk via hyperinsulinemia-linked endothelial inflammation [7].

5. Deep learning–guided oncometabolism: defining cancer-specific metabolic dependencies

Beyond diet–hormone interactions, advances in 2025 also highlighted how computational approaches may uncover metabolic vulnerabilities that are invisible to conventional experimental paradigms, particularly in cancer metabolism. A notable conceptual and methodological advance in 2025 was the application of deep learning to identify metabolic dependencies in cancer that are not readily apparent through conventional pathway analysis. The DeepMeta study by Wu et al. moved beyond descriptive associations by integrating large-scale transcriptomic data with curated metabolic networks to predict context-specific metabolic vulnerabilities across cancer types [8]. Rather than assuming uniform metabolic rewiring, the graph deep learning framework explicitly modeled how oncogenic states constrain metabolic flexibility. Using cancer gene-expression datasets from The Cancer Genome Atlas (TCGA), the investigators identified distinct metabolic nodes whose predicted dependency varied according to tumor lineage and genetic background. These predictions were experimentally validated, confirming that inhibition of specific metabolic pathways impaired cancer cell viability in a context-dependent manner. Notably, the analysis revealed recurrent reliance on pathways involved in nucleotide biosynthesis (particularly purine and pyrimidine metabolism), redox balance, and mitochondrial metabolism, underscoring the importance of metabolic homeostasis to malignant survival.

From a mechanistic standpoint, the study reframed oncometabolism as a problem of network constraint rather than global activation. Cancer cells were shown to depend on specific metabolic reactions not because they are universally upregulated, but because alternative routes are unavailable or incompatible with the oncogenic state. This perspective helps reconcile the long-standing observation that metabolic phenotypes differ widely between tumors despite shared driver mutations. A key example was the identification of CTNNB1 T41A-activating mutations, which encode β-catenin and drive oncogenesis in hepatocellular carcinoma and other tumor types, as creating a specific vulnerability to purine and pyrimidine metabolism inhibition. TCGA patients with the predicted pyrimidine metabolism dependency showed dramatically improved clinical response to chemotherapeutic drugs that block this pathway, confirming the clinical relevance of the computational predictions.

Beyond oncology, the broader significance of this work lies in its demonstration that artificial intelligence can function as a discovery engine in metabolism, capable of generating experimentally testable hypotheses at scale. By coupling deep learning with metabolic knowledge rather than treating metabolism as an abstract feature field, the study provides a template for how computational methods can advance mechanistic understanding. In 2025, this approach marked a decisive step toward data-driven, precision oncometabolism.

6. In vivo gene editing as metabolic therapy: from bespoke correction to one-time cardiometabolic intervention

The previous data-driven insights into metabolic constraint and dependency provided a conceptual backdrop for parallel advances aimed not merely at exploiting metabolic vulnerabilities, but at permanently reprogramming metabolic pathways through gene editing. Two reports in 2025 made gene editing feel less like a futuristic platform and more like an emerging therapeutic modality in metabolism, with the liver as the immediate proving ground.

The first report by Musunuru et al. describes a patient-specific CRISPR–Cas adenine base-editing therapy for neonatal-onset carbamoyl-phosphate synthetase 1 (CPS1) deficiency, an ultrarare and often fatal urea cycle disorder [9]. The team developed a bespoke adenine base editor (k-abe) delivered via lipid nanoparticles, customized to the patient's unique CPS1 Q335X variant. After rapid regulatory clearance, the infant received two intravenous infusions at approximately 7 and 8 months of age. In the 7 weeks following the initial infusion, the patient demonstrated clinically meaningful improvement: increased dietary protein tolerance and a 50 % reduction in nitrogen-scavenger medication, with no serious adverse events, even during intercurrent viral illnesses. Safety assessment included targeted off-target analysis, which showed low-level editing in a non-pathogenic intronic region in cell lines but not in primary hepatocytes, and no evidence of germline transmission was observed in analogous preclinical models using a different lipid nanoparticle gene-editing drug. The study highlights the feasibility of moving from diagnosis to a customized in vivo gene-editing therapy within six months, and suggests that similar approaches could be rapidly adapted for other hepatic inborn errors of metabolism.

The second study by Laffin et al. framed gene editing not as individualized rescue but as population-scale cardiometabolic prevention. ANGPTL3 is a hepatically produced protein that inhibits lipoprotein lipase and endothelial lipase, thereby regulating the metabolism of triglyceride-rich lipoproteins and LDL cholesterol. Naturally occurring loss-of-function variants in ANGPTL3 are associated with lifelong reductions in serum triglycerides and LDL cholesterol, and confer reduced risk of atherosclerotic cardiovascular disease. In a phase 1 trial of in vivo CRISPR–Cas9 editing of ANGPTL3 (CTX310) for cardiometabolic prevention, 15 adults with refractory hypercholesterolemia, hypertriglyceridemia, or mixed dyslipidemia received a single intravenous dose of CTX310, a lipid nanoparticle formulation encoding CRISPR–Cas9 mRNA and guide RNA targeting hepatic ANGPTL3 [10]. At the highest dose (0.8 mg/kg), mean reductions at 60 days were approximately 49 % in LDL cholesterol and 55 % in triglycerides, with no dose-limiting toxicities or serious adverse events attributable to the therapy. Infusion reactions and transient aminotransferase elevations were observed but resolved without sequelae. The trial demonstrates the potential for a one-time, durable gene-editing intervention to achieve substantial and simultaneous reductions in atherogenic lipoproteins, with regulatory guidance recommending up to 15 years of follow-up for long-term safety and efficacy.

Taken together, these studies mark a clear inflection point in 2025: gene editing in metabolism is no longer confined to the identification of “protective” variants. Instead, it is increasingly directed toward the deliberate engineering of durable metabolic states, whether through bespoke correction of catastrophic inborn errors or via one-time interventions designed to achieve lifelong lipid lowering.

7. Metabolome–microbiome interactions: biochemical stratification of impaired glucose regulation

As gene-editing studies illustrated the feasibility of durable molecular intervention, population-based omics studies emphasized that metabolic dysregulation in common diseases often arises from coordinated biochemical signatures rather than single genetic lesions. A large population-based study by Wu et al., published in 2025, addressed a central limitation of prior microbiome research in metabolism: the weak and inconsistent association between microbial composition and clinically meaningful metabolic phenotypes. The investigators analyzed two well-characterized Swedish cohorts comprising 1167 individuals, integrating shotgun metagenomics of the gut microbiome with targeted plasma metabolomics profiling of 502 circulating metabolites, alongside detailed metabolic phenotyping [11].

Impaired glucose regulation was associated with a broad and reproducible alteration of the circulating metabolome, involving branched-chain amino acids, aromatic amino acids, bile acid derivatives, lipid species, and microbial co-metabolites. Specifically, the microbiome–metabolome signature included elevated branched-chain and aromatic amino acids together with reduced levels of microbiome-derived metabolites, notably indole-3-propionic acid and hippurate. Approximately 140 metabolites showed statistically significant associations with microbial features, indicating that a substantial fraction of the dysregulated metabolome in dysglycemia is microbiome-linked. However, microbial diversity indices alone were only weakly predictive of glucose status, whereas metabolite-based signatures robustly discriminated normoglycemia from impaired glucose tolerance.

Importantly, the study included a structured lifestyle intervention with dietary modification and follow-up sampling. Improvements in glycemic indices were accompanied by normalization of specific metabolite clusters, particularly amino-acid–related and bile-acid–related metabolites, even when microbiome composition changed modestly. This temporal dissociation suggested that metabolic improvement tracks more closely with host–microbe metabolic output than with taxonomic shifts.

Mechanistically, the findings support a model in which the microbiome influences glucose metabolism through a limited set of circulating biochemical intermediates rather than through broad ecological restructuring. Clinically, the study positions plasma metabolomics as a superior integrative readout of microbiome–host interaction and a potentially scalable tool for metabolic risk stratification and monitoring of intervention response, advancing the field beyond descriptive microbiome associations.

8. Electron transport as a regulatory node of metabolism

At the cellular level, experimental work in 2025 showed that metabolic dysregulation can also arise from subtle perturbations in intracellular redox balance, even in the absence of overt structural damage. The experimental study by Goncalves et al. redefined the mitochondrial electron transport chain as an active regulator of hepatic metabolism, demonstrating that coenzyme Q (CoQ) redox imbalance induces reverse electron transport (RET) and excessive mitochondrial reactive oxygen species (mROS) production independently of ATP deficiency. Importantly, the study established that these redox disturbances drive hepatic steatosis and metabolic reprogramming through redox-sensitive regulation of metabolic enzymes and signaling pathways, rather than through structural mitochondrial damage [12].

The authors showed that impaired hepatic CoQ biosynthesis in obesity increases the CoQH₂/CoQ ratio, thereby promoting RET at complex I and triggering pathological mROS generation. These redox alterations were mechanistically linked to metabolic rewiring and lipid accumulation via modulation of redox-sensitive enzymatic and signaling networks, in the absence of overt mitochondrial structural injury. Restoration of CoQ redox balance attenuated RET-driven mROS production, identifying this process as a key driver of hepatic metabolic dysfunction in obesity and steatosis.

Collectively, these findings position CoQ redox status and RET-driven mROS as central mediators of hepatic metabolic pathology, shifting the conceptual framework from models of irreversible mitochondrial damage toward dynamic redox regulation. The work carries clear translational implications, identifying electron transport–linked pathways as potential therapeutic targets in dyslipidemia, insulin resistance, fatty liver disease, cardiometabolic disease, and diabetes. It also underscores the need to re-evaluate existing CoQ supplementation strategies and to develop formulations capable of achieving effective, tissue-targeted delivery. Moreover, this mechanism offers a plausible biological explanation for the increased risk of type 2 diabetes associated with statin therapy, potentially mediated by CoQ imbalance, and suggests that targeted hepatic CoQ replenishment merits further investigation.

In conclusion, the demonstration that excess hepatic mitochondrial ROS generation in obesity is site-specific highlights new therapeutic directions focused on limiting RET, restoring CoQ availability, or strategically combining both approaches.

9. Precision nutrition through sulfur amino acid manipulation: cysteine as a regulator of adipose thermogenesis

The above-mentioned findings reinforced the notion that metabolic state is highly sensitive to upstream nutrient availability, motivating renewed interest in how specific dietary components can program energy balance and thermogenic responses. A 2025 experimental study by Lee et al. identified cysteine as a previously underappreciated nutritional signal regulating adipose tissue thermogenesis and systemic energy balance [13]. Analysis of subcutaneous adipose tissue from participants in the CALERIE-II caloric restriction trial showed that sustained moderate caloric restriction in humans is accompanied by selective depletion of adipose cysteine, together with coordinated rewiring of sulfur amino acid metabolism. These observations provided the rationale for mechanistic investigation in experimental models, where systemic cysteine depletion was shown to induce rapid and profound weight loss (−30 %) driven primarily by fat mass reduction, without evidence of reduced activity, impaired protein synthesis, or overt tissue pathology.

Mechanistically, cysteine deprivation activated a thermogenic program in white adipose tissue, characterized by browning, increased energy expenditure, and enhanced fatty acid oxidation. This response was mediated by sympathetic nervous system activation and β3-adrenergic signaling, and was independent of both FGF21 and UCP1, and preserved under thermoneutral conditions, indicating engagement of non-canonical thermogenic pathways. Importantly, cysteine depletion reversed high-fat-diet–induced obesity in mice, improved glucose tolerance, lowered fasting glucose, and attenuated adipose tissue inflammation, linking sulfur amino acid availability directly to metabolic health.

From a translational perspective, the study supports the concept that dietary amino acid composition, not caloric load alone, may program thermogenic and metabolic responses. In humans, the observed reduction in adipose cysteine during caloric restriction suggests that targeted modulation of sulfur amino acid intake may contribute to the metabolic benefits of dietary interventions. While direct cysteine restriction cannot yet be recommended clinically, these findings provide a mechanistic framework for precision-nutrition strategies that selectively adjust sulfur amino acid availability, potentially mimicking aspects of caloric restriction–induced metabolic adaptation. Sulfur amino acid restriction (SAAR) has shown anti-obesity effects in both animal models and short-term human trials, with evidence of decreased leptin, increased ketone bodies, and upregulation of thermogenic and metabolic gene programs in white adipose tissue [14]. Moreover, observational data also link higher plasma cysteine levels to increased adiposity and metabolic syndrome risk [15]. Nevertheless, safety considerations are critical. Rapid and profound weight loss from severe cysteine depletion, as seen in animal models, may be associated with activation of stress responses, depletion of glutathione and coenzyme A, impaired mitochondrial function, and metabolic rewiring toward energetically inefficient pathways. Clinical translation requires careful dietary formulation to avoid malnutrition and to ensure adequate intake of essential nutrients. SAAR is not currently recommended as a routine clinical intervention, but ongoing research supports its potential as a precision-nutrition strategy for obesity and metabolic disease, pending further studies on long-term safety, feasibility, and efficacy. However, the study by Lee et al. establishes cysteine metabolism as a tractable and biologically important target for individualized nutritional intervention [13].

10. Artificial intelligence in metabolism: refining phenotyping and metabolic risk prediction

The increasing complexity of metabolic phenotypes highlighted by nutritional and mechanistic studies has necessitated analytical systems capable of capturing dynamic, individualized metabolic behavior. A 2025 study by Carletti et al. applying artificial intelligence to metabolic phenotyping addressed a critical limitation of conventional metabolic assessment: the reliance on static biomarkers such as fasting glucose and HbA1c to capture a fundamentally dynamic metabolic process [16]. The investigators analyzed data from 1137 individuals spanning normal glucose regulation, prediabetes, and type 2 diabetes, integrating continuous glucose monitoring (CGM) with lifestyle data, anthropometrics, and gut microbiome features. Rather than focusing on mean glucose levels, the analysis focused on glucose spike characteristics, including amplitude, frequency, and recovery kinetics.

Using machine-learning models trained on these multimodal inputs, the study identified reproducible glycemic response patterns that were not aligned with traditional diagnostic categories. Individuals with normal HbA1c values frequently exhibited exaggerated postprandial glucose spikes and delayed glucose recovery, phenotypes that clustered with metabolic features typical of early dysglycemia. Conversely, some participants with established diabetes showed relatively stable postprandial profiles despite elevated baseline markers. These AI-derived phenotypes were consistent within individuals over time, indicating stable underlying metabolic traits rather than short-term behavioral effects. Interestingly, model interpretability analyses revealed that prediction of adverse glycemic patterns was driven by combinations of features, including glucose dynamics, microbiome-associated variables, and behavioral factors, rather than by any single parameter. When compared with standard clinical metrics, the AI-based approach more accurately identified individuals at risk of metabolic deterioration, particularly in the transition from normoglycemia to prediabetes.

Overall, this study highlights that metabolic risk is encoded in temporal patterns rather than static averages, positioning artificial intelligence and CGM-derived dynamics as central tools for redefining metabolic phenotypes and supporting earlier identification of metabolic dysregulation at a stage when intervention is most effective.

11. Endogenous cyanide as a regulator of cellular metabolism

Finally, several experimental studies revisited fundamental biochemical assumptions, revealing that endogenous metabolites traditionally viewed as toxic may function as regulated metabolic signals under tightly controlled conditions. One of the most conceptually provocative experimental studies of 2025 challenged the entrenched view of cyanide as exclusively a metabolic toxin [17]. Using a combination of biochemical assays, subcellular imaging, proteomics and in vivo models, the investigators showed that low concentrations of cyanide are produced endogenously in mammalian cells and tissues, and exert regulatory effects on cellular metabolism.

The study by Zuhra et al. identified lysosomes as a major site of cyanide generation, with production stimulated by glycine under acidic conditions and dependent on peroxidase activity. Endogenously generated cyanide was shown to modulate mitochondrial electron transport and ATP production in a dose-dependent manner, enhancing bioenergetics at low concentrations while inhibiting respiration at higher levels. This bell-shaped dose–response effect mirrors the behavior of other endogenous gasotransmitters (NO, CO and H₂S) and suggests that cyanide participates in physiological signaling rather than representing a byproduct of metabolic stress. It also highlights the balance between cytoprotective and cytotoxic concentrations of diffusible mammalian gaseous mediators.

At the molecular level, cyanide induced S-cyanylation of cysteine residues on a broad range of metabolic proteins, altering enzymatic activity and influencing pathways involved in lipid oxidation, glycolysis, and redox regulation. In cellular and animal models, low-dose cyanide exposure conferred protection against hypoxic and ischemic stress, whereas excessive endogenous production, as observed in models of glycine metabolism disorders, was metabolically deleterious. Although the majority of the literature has traditionally framed cyanide primarily as a toxicant, it is important to note that cyanide is endogenously produced in several bacterial and plant species, where it exerts diverse regulatory functions, including roles in quorum sensing, biocontrol, germination, developmental processes and immune responses.

These findings redefine cyanide as a context-dependent metabolic regulator, expanding the repertoire of small-molecule signals involved in cellular energy control. In the landscape of metabolic research, this study exemplifies how revisiting fundamental biochemical assumptions cmay uncover previously unrecognized layers of metabolic regulation.

12. Conclusion

The metabolic advances reported in 2025 were unified not by scale, but by precision. Across clinical trials, mechanistic investigations and data-driven analyses, these studies underscored that meaningful metabolic modification is increasingly achievable when interventions are directed at defined pathways, dynamic phenotypes, or molecular control points.

Breakthroughs in obesity pharmacotherapy established new efficacy benchmarks [[18], [19], [20]], while gene-editing studies revealed the feasibility of durable metabolic reprogramming in humans. At the same time, mechanistic work redefined the roles of mitochondrial redox balance, nutrient-derived signals and endogenous gaseous mediators in metabolic regulation. Multi-omics and artificial intelligence approaches further showed that metabolic risk and response are encoded in coordinated biochemical and temporal patterns rather than static measurements [[21], [22], [23]].

In conclusion, these studies position metabolism as a biologically tractable system that can be characterized, stratified and modified with increasing precision. The challenge ahead is to translate this precision responsibly while balancing efficacy, safety, durability, and accessibility, and integrating clinical, molecular, and computational insights into strategies for metabolic disease prevention and treatment.

CRediT authorship contribution statement

Maria Dalamaga: Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Project administration, Supervision, Writing – original draft, Writing – review & editing. Junli Liu: Writing – review & editing.

Funding information

None.

Conflict of interest

None.