Clear cell renal cell carcinoma (ccRCC) is the most common and aggressive form of kidney cancer. Despite the introduction of targeted therapies and immune checkpoint inhibitors, many patients do not respond well, and outcomes vary widely. Better tools for predicting who will respond to which treatment are urgently needed to improve patient care.
Palmitoylation is a type of chemical modification that cells use to control where proteins go and what they do. It works by attaching a fatty acid molecule called palmitic acid to specific proteins, which changes how those proteins interact with cell membranes and other cellular components. This modification is reversible and can be switched on or off depending on what the cell needs.
In cancer, abnormal palmitoylation has emerged as an important driver of tumor growth, spread, and resistance to treatment. The proteins responsible for adding the palmitic acid tag belong to the ZDHHC family, which includes 24 distinct enzymes (ZDHHC1 through ZDHHC24). When these enzymes are dysregulated in cancer, they can change the behavior of key proteins involved in cell division, migration, and survival.
Despite growing evidence that palmitoylation plays a critical role in cancer biology, its specific impact on ccRCC prognosis and response to immunotherapy had not been well studied. This study addressed that gap by combining large-scale genomic data analysis, machine learning, and laboratory experiments to uncover how palmitoylation shapes kidney cancer outcomes and identify a key gene called ZDHHC18 as a promising therapeutic target.
The researchers integrated five large publicly available ccRCC gene expression datasets, covering a total of more than 1,000 patients from the TCGA, ICGC, and other international databases. Before analyzing the data, batch effect correction was applied using the Combat algorithm to remove technical differences between studies, ensuring that the combined dataset reflected true biological differences rather than laboratory artifacts.
Twenty-five key palmitoylation-related genes were identified by cross-referencing multiple specialized databases. To find which genes were most important for predicting patient survival, the researchers tested 101 different combinations of 10 machine learning algorithms, including Lasso regression, Random Survival Forest (RSF), CoxBoost, Gradient Boosting, and others. Each combination was evaluated using the concordance index (C-index), a measure of how accurately the model predicts who will survive longer.
The combination of Lasso regression followed by Random Survival Forest produced the highest C-index across all three datasets (training set and two independent validation cohorts) and was selected as the final prognostic model. Lasso regression first narrowed down the gene list to the 10 most relevant palmitoylation genes, which were then used in the RSF analysis to calculate a risk score for each patient.
The study also used single-cell RNA sequencing data from five separate datasets to examine which specific cell types within the tumor were expressing the key genes identified by the model. Additionally, laboratory experiments using two kidney cancer cell lines (786-O and Caki-1) were conducted to directly test the function of the top gene, ZDHHC18, by suppressing its activity and observing the effects on cancer cell behavior.
Analysis of palmitoylation levels across the combined patient cohort revealed that ccRCC tumor samples had significantly lower palmitoylation activity compared to normal kidney tissue. Importantly, patients with different levels of palmitoylation had significantly different survival outcomes, confirming that this molecular process is clinically meaningful in kidney cancer.
From the 25 palmitoylation-related genes studied, Lasso regression identified 10 key genes. RSF analysis of these 10 genes found that ZDHHC18 had the highest variable importance score, meaning it contributed most strongly to predicting patient survival. Based on the combined risk score from these genes, patients were divided into high-risk and low-risk groups, with the high-risk group showing significantly shorter survival (log-rank p less than 0.001) in the training dataset and confirmed in independent validation cohorts.
The predictive accuracy of the model was excellent, with an area under the curve (AUC) of 0.99 in the training set and 0.75 and 0.63 in the two validation datasets at 1 year. The risk score remained a significant independent predictor of survival even after accounting for clinical factors such as age, cancer grade, and tumor-node-metastasis (TNM) staging, confirming that palmitoylation-related gene activity provides prognostic information beyond what standard clinical measures capture.
High-risk patients showed elevated activity of cancer-promoting pathways including PI3K/AKT/mTOR signaling, heme metabolism, and fatty acid metabolism, which are all associated with aggressive tumor behavior. They also showed higher tumor mutational burden (TMB), greater chromosomal instability (more copy number variations), and higher tumor stemness scores, all indicators of more dangerous disease.
To confirm that ZDHHC18 plays a direct functional role in kidney cancer, the researchers used gene silencing technology (stable knockdown) to reduce ZDHHC18 activity in two ccRCC cell lines. This allowed them to observe what happens to cancer cells when ZDHHC18 is turned down, mimicking the therapeutic effect of blocking this gene.
When ZDHHC18 was silenced, kidney cancer cells showed a significant reduction in their ability to proliferate (grow and divide). This was confirmed by three independent assays: CCK-8 cell viability testing, colony formation assays measuring long-term growth, and EdU assays directly measuring DNA replication. Fewer cells were dividing, and fewer new colonies formed, indicating that ZDHHC18 is necessary for active cancer cell growth.
Silencing ZDHHC18 also significantly impaired the cancer cells' ability to migrate and invade surrounding tissue, as shown by Transwell assays. This is particularly important because cancer invasion is the first step toward metastatic spread. If ZDHHC18 is needed for cancer cells to invade, then blocking it could potentially prevent kidney cancer from spreading to other organs.
Crucially, silencing ZDHHC18 in normal kidney cells (HK-2 cell line) had no significant effect on proliferation or migration. This specificity suggests that ZDHHC18 is selectively important for cancer cell biology and is not essential for normal kidney function, a promising characteristic for any potential therapeutic target since it reduces the risk of off-target side effects.
One of the most clinically valuable findings of this study is the link between the palmitoylation-based risk score and immunotherapy response prediction. Using the TIDE algorithm and the independent Submap analysis method, the study found that low-risk patients were significantly more likely to respond favorably to immune checkpoint inhibitor therapy, while high-risk patients showed signs of immune dysfunction that may reduce treatment effectiveness.
Immune profiling analysis showed that high-risk patients had higher immune scores overall (more immune cells present in their tumors) but paradoxically were less likely to respond to immunotherapy. This pattern is known as immune exclusion or immune dysfunction, where the immune system is present but functionally suppressed by the tumor. This type of immune environment is particularly challenging to treat with standard checkpoint inhibitors.
The risk score was also positively correlated with the activity of immune-related pathways including PD-1 and CTLA-4 blockade pathways, suggesting that the palmitoylation-related model captures information relevant to how tumors interact with the immune system. Patients classified as high risk by the palmitoylation model may need combination immunotherapy strategies or alternative treatment approaches beyond standard checkpoint inhibition.
For patients with ccRCC, these findings suggest that measuring ZDHHC18 expression and calculating a palmitoylation-based risk score from tumor biopsy samples could one day help oncologists decide which patients are most likely to benefit from immunotherapy versus those who may need a different treatment approach from the start, reducing the time patients spend on ineffective treatments.
This study establishes palmitoylation-related genes as a new category of prognostic biomarkers in ccRCC. By testing 101 machine learning algorithm combinations and validating the best model across multiple independent patient cohorts, the researchers built a robust and clinically meaningful tool for predicting which kidney cancer patients face the greatest risk of poor outcomes.
The identification of ZDHHC18 as both the most important prognostic gene and a validated functional driver of kidney cancer cell growth and invasion is a significant finding. ZDHHC18 overexpression in tumor tissue compared to normal kidney tissue makes it a potentially measurable biomarker in clinical biopsy samples, and its cancer-specific functional role makes it an attractive therapeutic target for future drug development.
The combination of computational analysis and laboratory validation used in this study is a model for translational cancer research. By first identifying candidate targets through large-scale genomic analysis and then confirming their function in cell line experiments, the researchers have built a stronger foundation for future clinical applications than either approach could provide alone.
Future research directions include developing inhibitors specifically targeting ZDHHC18 palmitoylation activity, testing whether restoring normal palmitoylation levels in high-risk tumors can improve immunotherapy response, and validating the risk score prospectively in clinical trials. These steps will determine whether the palmitoylation model can be incorporated into routine clinical practice to guide personalized treatment decisions for kidney cancer patients.