Cancer cells fundamentally rewire their metabolism to support rapid growth - a phenomenon known as the Warburg effect, where tumors preferentially use glycolysis (sugar breakdown) even when oxygen is available. These metabolic changes are not just side effects of cancer; they actively drive tumor progression and influence how tumors respond to treatment.
Current endometrial cancer (EC) classification systems focus primarily on molecular features like POLE mutations, microsatellite instability, and TP53 status. While powerful, these systems do not fully capture the metabolic heterogeneity of EC - the wide variation in how different tumors fuel their growth - which may have independent implications for prognosis and treatment response.
This study performed a comprehensive multi-omics analysis - simultaneously examining RNA expression, protein levels, and metabolite concentrations - to identify distinct metabolic subtypes within endometrial cancer. The goal was to create a classification system based on metabolism that could complement existing molecular classifications and better guide treatment decisions.
The researchers analyzed data from The Cancer Genome Atlas (TCGA), a large public database containing molecular data from thousands of cancer patients. For EC patients, they examined RNA-seq (which genes are active), proteomics (which proteins are present), and metabolomics (which metabolites are present) data in concert.
Using consensus clustering - a machine learning approach that groups patients based on the similarity of their metabolic gene expression patterns - the researchers identified distinct subgroups. This unsupervised approach lets the data define natural groupings rather than imposing predetermined categories.
A 13-gene hub classifier was then developed using LASSO regression (a regularization technique that selects the most informative features) to enable practical assignment of new patients to metabolic subgroups. This classifier distills the complex multi-omics landscape into a small, measurable gene panel suitable for clinical testing.
The analysis revealed three distinct metabolism pathway-based subgroups (MPS1, MPS2, MPS3). Each subgroup showed a characteristic pattern of metabolic pathway activation that distinguished it from the others, reflecting fundamentally different ways each tumor type fuels its growth.
MPS3 showed the worst prognosis, with a hazard ratio of 2.22 for overall survival compared to other groups - meaning MPS3 patients had more than double the risk of death during the study period. MPS3 tumors were characterized by high activation of multiple metabolic pathways and corresponded to more aggressive tumor histology.
MPS2 showed the best predicted response to immunotherapy, based on immune infiltration scores and expression of immune checkpoint markers. This suggests that metabolic subtyping could help identify EC patients most likely to benefit from checkpoint inhibitors like pembrolizumab, which are increasingly used in recurrent EC treatment.
Two metabolic pathways were consistently upregulated across EC tumors compared to normal endometrium: glycolysis/gluconeogenesis (the core sugar-metabolism pathway associated with the Warburg effect) and folate biosynthesis (important for DNA synthesis and one-carbon metabolism). These represent universal metabolic adaptations in EC.
The three subtypes differed in their activation of additional pathways beyond this shared core. MPS3 showed the broadest metabolic reprogramming, with upregulation of pathways including amino acid metabolism, lipid synthesis, and nucleotide biosynthesis - consistent with the higher energy demands of rapidly proliferating aggressive tumors.
LASSO regression works by penalizing the inclusion of too many variables, forcing the model to select only the most informative genes. From thousands of metabolism-related genes, this approach identified a 13-gene panel that captures the essential features of each metabolic subtype - a practical distillation enabling routine clinical testing rather than expensive full-genome assays.
The identification of MPS2 as the subgroup most likely to respond to immune checkpoint inhibitors has direct clinical implications. Checkpoint inhibitors work by releasing immune cells from suppression, allowing them to attack tumors. MPS2's immune-favorable microenvironment - reflected in higher T-cell infiltration and PD-L1 expression - creates conditions where these drugs can be most effective.
For MPS3 patients with poor prognosis, the detailed metabolic profile identifies potential vulnerabilities. Drugs targeting glycolysis or folate metabolism (like metformin, which affects glucose metabolism, or antifolates like methotrexate) might be particularly effective for this subgroup. Metabolic subtyping could guide allocation of these existing drugs to patients most likely to respond.
The 13-gene classifier needs prospective validation in independent patient cohorts before clinical adoption. However, the relatively small panel size means it could potentially be incorporated into existing targeted sequencing panels or gene expression assays used in standard oncology practice without requiring new specialized testing infrastructure.
This study establishes that endometrial cancer contains distinct metabolic subtypes with different clinical outcomes and treatment implications - a dimension of heterogeneity not captured by existing molecular classification systems. The MPS classification adds complementary information to the established POLE/MSI/TP53 framework.
The integration of three omics layers (transcriptomics, proteomics, metabolomics) provided more robust subtype identification than any single layer alone, demonstrating the value of multi-omics approaches for understanding cancer biology. Proteomics and metabolomics data, which are less commonly collected than gene expression data, contributed unique information about actual cellular function beyond what genes are active.
Future directions include combining metabolic subtyping with established molecular classifications to create a more comprehensive EC stratification system, and testing whether metabolic subtype-guided treatment allocation actually improves patient outcomes in prospective clinical trials. The biological insights from this study also reveal novel drug targets within altered metabolic pathways.