VHL gene mutation or inactivation is one of the most common molecular characteristics of ccRCC. The loss of VHL function leads to the stabilisation of HIF-1α and HIF-2α (Cancer Genome Atlas Research Network, 2013; Jaakkola et al., 2001; Bao et al., 2019). HIF is an oxygen-sensitive transcription factor that helps tumor cells adapt to a hypoxic environment by upregulating a series of genes. Specifically, HIF-1/2α promotes glucose uptake in tumor cells, acidification of the tumor environment, and the formation of new blood vessels by upregulating genes such as GLUT1, CA9, and VEGF, and supports tumor growth (Zh et al., 2023; Almanzar et al., 2025; Reinfeld et al., 2022). In addition, HIF-1/2α interacts closely with mTOR, MYC, PI3K/AKT, and other signalling pathways, jointly regulating multiple levels of tumor metabolism, including glycolysis, fatty acid synthesis, and amino acid metabolism (Badoiu et al., 2023; Yecies and Manning, 2011; Li et al., 2020).

For the first-in-class HIF-2α inhibitor belzutifan, the process has been very rapid, first a flurry of studies, and now clinical trials. The latest clinical data show that belzutifan can effectively improve progression-free survival (PFS) and symptom relief in patients with ccRCC by inhibiting HIF-2α activity, especially when combined with other targeted drugs (Jonasch et al., 2021; Choueiri et al., 2024; Song et al., 2024).

A key aspect of metabolic reprogramming is the Warburg effect: in the presence of oxygen, tumor cells preferentially metabolise glucose through glycolysis instead of oxidative phosphorylation to generate energy. ccRCC cells enhance the glycolysis pathway to facilitate glucose uptake, lower the pH in tumor cells by lactic acid accumulation, and support a beneficial environment for their growth (Zh et al., 2023; Yang et al., 2023).

In ccRCC, tumor cells suppress the TCA cycle and modify mitochondria to adapt to heightened metabolic demands. This metabolic transformation is intimately associated with immunosuppression, drug resistance, and metastasis. Research suggests that the Warburg effect not only supports tumor cell growth but also suppresses immune cell function through changes in the TME metabolic milieu (e.g., lactic acid and hypoxia), thereby increasing resistance to immune checkpoint inhibitors (ICI) and chemotherapeutic agents (Zh et al., 2023; Yang et al., 2025).

Lipid metabolism is a key component of reprogramming ccRCC metabolism. Cancer cells synthesize, store, and mobilize lipids to favour the formation of cell membranes and energy reserves in tumor cells. Lipid droplet and cholesterol ester accumulation are hallmarks of ccRCC, and tumor cells express critical enzymes driving lipogenesis, including fatty acid synthase (FASN), unsaturated fatty acid synthase (SCD), and long-chain lipoacyl-CoA synthase (ACSL) (Deng et al. 2025; Klasson et al., 2022; Li et al., 2023).

Lipid metabolism is closely linked to iron-dependent cell death (ferroptosis). The buildup of lipid peroxides triggers ferroptotic cell death, while unsaturated fatty acids and the GPX4/xCT pathways regulate this lipid-peroxide–driven process. Tumor cells cope with metabolic stress and determine their tolerance to treatment by regulating the activity of these pathways (Zh et al., 2023; Zhou et al., 2024).

Metabolism of amino acids is vital for the regulation of redox homeostasis, the supply of NADPH, and the synthesis of nucleotides in tumor cells. ccRCC cells show a strong dependency on glutamine and serine/one-carbon metabolism. Through these pathways, tumor cells can maintain their antioxidant ability and meet the needs of rapid proliferation (Zhang et al., 2025a; Kaushik et al., 2022; Lyu et al., 2025). In recent years, metabolic inhibition and synthetic lethal strategies have emerged as a new direction in tumor treatment. For example, drugs targeting glutamine metabolism, amino acid transport, and one-carbon metabolism have shown promising efficacy in early clinical trials, especially when combined with other therapies (Zh et al., 2023; Jin et al., 2023; Ren et al., 2024).

Single-cell and spatialomics techniques revealed high metabolic heterogeneity in ccRCC, reflecting differences across tumor regions and cell subgroups (Shi et al., 2025). In particular, there is a line called “de-lipid-droplet cell differentiation (DCCD)” in the tumor. These tumor cells have fewer lipid droplets but exhibit strong nutrient uptake, rapid proliferation, and a poor prognosis (Hu et al., 2024; Yang et al., 2024). The heterogeneity of metabolic states offers new opportunities for individualised treatment strategies and stratified approaches to overcome tumor drug resistance and immune escape. Table 1 further illustrates the key characteristics of each metabolic pathway and its therapeutic significance.

The lipid-rich, hypoxic metabolic ecology of ccRCC will systematically “transform” the metabolic state and function of immune cells in the TME, which is a key driver of immunotherapy drug resistance (Niu et al., 2025a). The metabolic reprogramming of myeloid immune cells is one of the driving forces shaping the immunosuppressive microenvironment. Tumor-associated macrophages (TAM) tend to absorb a large amount of oxidative lipids released by tumor cells in ccRCC through scavenger receptors (such as CD36). These lipids drive mitochondrial β-oxidation in intracellular cells, promote their polarisation to an immunosuppressive M2 phenotype, and secrete IL-10, TGF-β, and other inhibitory cytokines (Hu et al., 2024; Simeth et al., 2025). In other urinary system tumors, transcriptomic studies centred on macrophages have revealed an immune-regulatory mechanism driven by lipid metabolism, which also supports the existence of similar metabolic-immune mechanisms in ccRCC (Wang et al., 2023; Wang et al., 2024). Myeloid-derived suppressor cells (MDSCs) use fatty acid metabolism to maintain their survival and inhibitory functions, and directly inhibit T cell proliferation and activity by consuming essential amino acids, such as arginine and cysteine, in the microenvironment (Zhang et al., 2025b).

In the T cell compartment, the function of CD8⁺ T cells is suppressed by multiple mechanisms, including lactic acid accumulation: the high lactic acid environment produced by tumor glycolysis directly inhibits mTOR signalling and T cell function (Brand et al., 2016; Watson et al., 2021). Lipid coercion: Excessive lipids in the TME lead to mitochondrial dysfunction in T cells and upregulate markers of T cell depletion, such as PD-1 (Fadini, 2020; Lin et al., 2020). Nutritional competition: The competition between tumor cells for glucose and glutamine causes T cells to “lack metabolism” (Cha et al., 2015; Ho et al., 2015).

These metabolic mechanisms together lead to T cell depletion and apoptosis, significantly weakening the efficacy of immune checkpoint inhibitors (ICI). It is worth noting that tumor lipid metabolism is systematically negatively correlated with the infiltration density and functional status of CD8⁺ T cells (Simeth et al., 2025; Zhang et al., 2025b). Therefore, targeting abnormal lipid metabolism in tumor and medullary cells, or relieving T-cell metabolic stress through metabolic regulators (such as LDHA inhibitors and PPARα agonists), has become a promising strategy to improve the efficacy of ICI.

Non-invasive “metabolic fingerprints” provide a new tool for the clinical management of ccRCC. In body fluid metabolomics, plasma- or urine-based metabolite analysis has shown great potential. Using mass spectrometry and machine learning, markers including succinic acid, acylcarnitine, and specific phospholipids have been identified. These molecular characteristics can distinguish ccRCC from healthy individuals with high accuracy, and a multi-centre cohort has validated their reliability for early diagnosis and risk stratification (Ho et al., 2015; Huang et al., 2025; Ossolińska et al., 2025). At the level of image integration, the radiopathomics framework deeply correlates CT image information with molecular metabolic phenotype. The core finding is a significant negative correlation between the tumor lipid attenuation (HU value) in CT imaging and the expression level of the lipid droplet marker PLIN2 in the tissue (Deng et al., 2025; Wang, 2023). This relationship enables clinicians to infer a tumor’s lipid metabolic status noninvasively using routine CT imaging.

At present, artificial intelligence is further promoting the transformation of this field. Using a training model to decode complex features (texture, shape, etc.) in CT images, key metabolic phenotypes, such as PLIN2 expression and ferroptosis sensitivity, can be predicted, and patients with “high-risk” metabolic characteristics (such as DCCD tendency) can be identified before treatment (Deng et al., 2025). In the end, the combination of “body fluid metabolism spectrum + image lipid reading” is being built into a powerful, non-invasive diagnostic support system.

Hypoxic axis: With the emergence of the HIF-2α inhibitor belzutifan, the hypoxia-driven metabolic–angiogenic program can be pharmacologically inhibited for the first time. The phase III LITESPARK-005 trial demonstrated that in patients with advanced ccRCC previously treated with ICIs and anti-angiogenic therapy, belzutifan improved PFS and ORR compared with everolimus, without new safety signals (Choueiri et al., 2024; Powles et al., 2025). Lipid metabolic reprogramming is tightly linked to ferroptosis. The system x_c⁻ (xCT/SLC7A11)–GSH–GPX4 axis represents a key brake on ferroptosis. Emerging evidence supports strategies that suppress lipid storage and/or promote lipid peroxidation to induce ferroptosis, potentially in combination with immunotherapy and anti-angiogenic therapy to broaden the therapeutic window (Deng et al., 2025; He et al., 2024).

Amino acids and one-carbon metabolism: ccRCC’s dependence on glutamine and one-carbon metabolism provides an entry point for synthetic death; at the mechanism and early clinical level, targets such as GLS/transporters show the potential to reverse drug resistance and improve immune response, and be alert to compensatory pathways and systemic toxicity (Wang, 2023).

Metabolic remodeling, vascular “normalization,” and T-cell infiltration can form a synergistic cascade: inhibiting HIF–VEGF signaling and dysregulated lipid metabolism may reduce myeloid-mediated immunosuppression, improve perfusion and oxygenation, and enhance CD8+ T-cell effector function, thereby providing a metabolic rationale for synergy between TKIs/VEGF inhibitors and ICIs. On the other hand, the strategy of lipid suppression or induction of lipid peroxidation + ICI is expected to overcome immune tolerance by improving antigen presentation and immunoinflammatory microenvironment (Deng et al., 2025; Zhang et al., 2025b).

Multi-omics stratification is being formed: integrated analysis of transcriptome–lipidome–immune infiltration reveals that there is a coupling relationship between the strength of lipid metabolic gene pathway and CD8⁺ T cell infiltration/function, and survival outcome, which can be used to guide the priority of “metabolism-immune” combination (Simeth et al., 2025); radiopathomics, represented by the coupling of ADFP/PLIN2-CT HU, provides non-invasive lipid droplet/lipid metabolism readings (Deng et al., 2025); the combination of fluid metabolism fingerprints (plasma/urine) and AI models provides deployable companion diagnostics for early screening, efficacy prediction and follow-up. Tools (Ho et al., 2015; Ossolińska et al., 2025).

The plasticity of tumor metabolism makes it diverse in different microenvironments. Especially in ccRCC, the metabolic interactions between tumor cells and host organs (such as the liver and adipose tissue) remain poorly understood. Research shows that the metabolic characteristics of ccRCC are also closely linked to the host’s systemic metabolic status (such as obesity-related lipid mobilisation and diabetes-related hyperglycaemia) (Greco et al., 2024; Venkatesh et al., 2023). For example, the free fatty acids released by adipose tissue in obese patients may be taken up by ccRCC cells and used to synthesize lipid droplets or as an energy source, potentially affecting tumor aggressiveness and modifying therapeutic response (Yun et al., 2025; Niu et al., 2025b). Therefore, it is crucial to develop effective, individualised treatment strategies to understand metabolic interactions within the “tumor-host” ecosystem beyond the tumor itself.

Metabolic reprogramming enables tumor cells to survive in harsh environments and may also improve treatment tolerance. For example, tumor cells may avoid the effect of drugs by changing their metabolic pathways, such as surviving by upregulating alternative amino acid metabolic pathways after using HIF-2α inhibitors, or resisting the oxidative stress caused by treatment by enhancing specific metabolic pathways (such as the synthesis of antioxidant glutathione) (Peng et al., 2025; Gavi et al., 2025; Li et al., 2024). This metabolic “detour” or redundancy is an essential mechanism for the development of acquired drug resistance, which also explains why single-targeted metabolic therapy is often effective (Li et al., 2024; Nho et al., 2025). The metabolic adaptive mechanism has introduced new challenges to treatment, and there is an urgent need to develop a joint treatment strategy that targets multiple key metabolic nodes simultaneously.

There are still two core obstacles to translating findings on metabolic reprogramming into clinical practice. One is the lack of evidence level: at present, the scale and positive results of randomised controlled studies on metabolic targets are limited (for example, telaglenastat combined treatment fails to improve the outcome), and the mainstream guidelines also place it in the exploration stage (Tannir et al., 2022; Powles et al., 2024). The second is cross-centre repeatability and data governance: the metabolic group/lipid group needs to follow the sample processing and minimum report set of MSI and best practises, and the imaging end needs to unify the collection, reconstruction and feature definition according to the consensus of IBSI and machine learning to reduce batch differences and improve comparability (Sumner et al., 2007; Köfeler et al., 2021; Unterrainer et al., 2022) At the level of data co-construction, the FAIR principle should be implemented and federal learning should be adopted to achieve multi-institutional modelling under the premise of “not leaving the hospital” (Wilkinson et al., 2016; Sheller et al., 2020); at the same time, the study of tumor markers and diagnostic accuracy should follow REMARK/STARD and other reporting norms to improve external verifiability and Clinical promotion value (Sauerbrei et al., 2018; Bossuyt et al., 2015). Within the above-mentioned standardisation and governance framework, rigorously designed prospective, multi-centre clinical trials can confirm the real-world clinical value of metabolic targeting strategies.

Future research should focus on developing precise treatment strategies tailored to metabolic characteristics. For example, by analysing the metabolic phenotypes of tumor cells, such as DCCD status, lipid droplet accumulation, ferroptosis sensitivity, and HIF activity, an individualised treatment decision tree can be constructed (Hu et al., 2024; Dong et al., 2025). The preliminary decision-making framework could prioritize testing the “induced ferroptosis + ICI” scheme for “fat droplet enrichment/ferroptosis susceptibility” patients. In contrast, for “DCCD/fat droplet deficiency” patients, a combination of “HIF-2α inhibitor + targeted amino acid metabolism” may be required (Deng et al., 2025; Yang et al., 2025; He et al., 2024). These metabolic characteristics can serve as dynamic biomarkers to predict treatment response and drug resistance, providing a basis for truly individualised treatment.

The integration of artificial intelligence (AI) and multi-group data provides new perspectives and tools for the metabolic research of ccRCC. By combining genomics, transcriptomics, metabolomics, and imaging data, AI can help identify new metabolic markers and therapeutic targets (Dai et al., 2025; Zheng et al., 2025a; Chen et al., 2023). More importantly, AI can integrate heterogeneous data to build a comprehensive model that predicts a patient’s prognosis and treatment response (Chen et al., 2023; Zheng et al., 2025b). The progress of imaging metabolism enables the non-invasive, real-time monitoring of tumor metabolism, providing dynamic support for clinical decision-making. In the future, the integration of AI-driven multi-omics and imaging metabolism will be explicitly reflected in the development of “adaptive clinical trial design” and “intelligent companion diagnostics system”, thereby systematically promoting the clinical translation of ccRCC metabolism research. As shown in Figure 1, the plasticity of the tumor-host interaction and the metabolic reprogramming of metabolic pathways provide profound insights into the progression of ccRCC. In particular, in tumor-host metabolic interactions, host factors such as obesity and diabetes significantly affect tumor invasiveness and response to treatment. The metabolic adaptive bypass reveals how tumor cells escape drug treatment through metabolic reprogramming, providing new ideas for the development of precise treatment strategies.

Combined imaging–multi-omics markers can identify metabolic subtypes, such as “high glycolysis/low lipid-droplet load (including DCCD)/glutamine dependence,” which inform cohort stratification and efficacy prediction. The index is the objective remission rate of post-stratified first-line or combined treatment/progressive survival, which is significantly improved, and its repeatability has been reported (Fresnedo et al., 2025; Deng et al., 2025; Greco et al., 2024).

Starting in phase II, a threshold-locking companion diagnostic strategy was developed in parallel, and a basket/umbrella design with a medium-term suspension rule was implemented to enable rapid validation of the metabolic target. The indicators were the preset endpoint achievement rate, medium-term decision-making accuracy, and cross-centre consistency (Ho et al., 2015; Greco et al., 2024; Knott et al., 2018).

Collection, quality control, and sharing are carried out in accordance with MSI/IBSI and FAIR specifications. The indicators are the same effect direction of the cross-centre reproduction experiment, the performance difference within ≤ preset range, and the publicly available analysis process is formed (Greco et al., 2024; Choueiri et al., 2021; Zwanenburg et al., 2020).