Pancreatic ductal adenocarcinoma (PDAC) is one of the deadliest cancers, with a 5-year survival rate of only about 10%. A key reason for this is the tumor's unique microenvironment, which is characterized by very low oxygen levels (hypoxia) and dense fibrous tissue that blocks immune cells from entering.
When cancer cells are starved of oxygen, they switch to an alternative energy pathway that produces large amounts of lactate. This lactate can attach to proteins in a process called lactylation, altering how genes are expressed and potentially making the cancer more aggressive. This study explored the connection between hypoxia and lactylation to find new biomarkers and treatment targets.
Researchers started by identifying genes linked to both hypoxia and lactylation in PDAC patients from public databases. They applied three complementary machine learning algorithms — LASSO regression, XGBoost, and Random Forest — to narrow down thousands of candidate genes to the most predictive ones.
They also used single-cell RNA sequencing analysis to understand which cell types within the tumor expressed these genes, and performed CERES scoring to identify potential therapeutic targets. Laboratory experiments using cell assays confirmed the functional role of the top candidate gene, CENPA, in cancer cell proliferation and movement.
Using the identified genes, researchers classified PDAC patients into two biological subtypes. Subtype A was associated with significantly worse survival compared to Subtype B. Gene pathway analysis showed that Subtype A tumors had higher activation of cell cycle, epithelial-mesenchymal transition (EMT), and hypoxia pathways, all of which contribute to cancer spread and treatment resistance.
A risk score called the NRMS-like model was built from the top genes and validated in multiple independent patient cohorts. Patients with high risk scores consistently had worse overall survival. The high-risk group also showed a more immunosuppressive tumor environment with fewer active immune cells, explaining why these tumors tend to resist immunotherapy.
Among all the genes analyzed, CENPA emerged as a standout. It was significantly overexpressed in pancreatic cancer patients who also had new-onset diabetes, a known risk factor and early warning sign of pancreatic cancer. High CENPA expression correlated with faster tumor growth and greater immune suppression.
Laboratory experiments showed that reducing CENPA expression in pancreatic cancer cells slowed their proliferation and reduced their ability to migrate and invade surrounding tissue. These results position CENPA as a potential drug target — inhibiting it could simultaneously slow tumor growth and potentially make the immune system more effective.
One of the most clinically valuable findings was that the risk score could predict a patient's likely response to immunotherapy. Patients with a low risk score had more active immune cells and better outcomes when treated with immune checkpoint inhibitors. High-risk patients showed immune checkpoint gene patterns suggesting they would be less likely to respond.
This means oncologists may eventually be able to use a genetic risk score derived from a patient's tumor to decide whether to try immunotherapy or pursue alternative strategies. The ability to predict immunotherapy response before starting treatment could spare patients from ineffective treatments and help identify those who would truly benefit.
This study established a novel connection between hypoxia and lactylation in pancreatic cancer, creating a gene signature that predicts prognosis and immunotherapy response. The risk score was validated across multiple independent datasets, supporting its potential clinical usefulness.
The identification of CENPA as a therapeutic target opens a new direction for drug development. Future studies should explore whether drugs targeting CENPA or the broader hypoxia-lactylation pathway could improve outcomes for patients with this difficult cancer, particularly those with co-existing new-onset diabetes.