Metabolic Subtyping Predicts Immunotherapy Response in Nasopharyngeal Carcinoma

Clinical Summary:

• Design/Population: A biomarker study of 407 patients with locoregionally advanced nasopharyngeal carcinoma enrolled in the phase 3 CONTINUUM and DIPPER trials used RNA sequencing and metabolic gene clustering to define metabolic subtypes and evaluate benefit from chemoradiotherapy with or without PD-1 inhibitors.
• Key Outcomes: Three metabolic subtypes were identified, with only the MS1 subtype showing significant improvement in event-free survival with the addition of PD-1 blockade to chemoradiotherapy.
• Clinical Relevance: Metabolic subtyping may serve as a predictive biomarker to guide immunotherapy use, enabling more precise, individualized treatment strategies.

Sai-Wei Huang, MD, Sun Yat-sen University Cancer Center, Guangzhou, China, discusses biomarker findings from the phase 3 CONTINUUM and DIPPER trials which aimed to identify metabolic subtypes that predict response to PD-1-based chemoradiotherapy in patients with locoregionally advanced nasopharyngeal carcinoma. 

Results demonstrated that only the MS1 subtype derived significant event-free survival benefit from adding immunotherapy, while MS2 and MS3 patients showed limited benefit, supporting a biomarker-driven approach to personalize treatment and avoid unnecessary toxicity. 

Transcript:

Hi there, good morning, everybody. Thank you for the opportunity to talk about our group's recent work. I'm Sai-Wei Huang, a doctoral student at Sun Yat-sen University Cancer Center. Today, I'm going to discuss a biomarker study that we recently published in the Journal of Clinical Oncology.

First, let me talk about nasopharyngeal carcinoma (NPC). For many doctors in the West, NPC may be a rare disease that represents a significant health burden, especially in China, which accounts for 47% of the world's cases. NPC is located near the skull base and major vessels and nerves so surgery isn't suitable. Over 99% of cases are non-keratinizing carcinoma and are radiosensitive so radiotherapy is the mainstay treatment. Adding chemotherapy further reduces metastasis, thus above 20% of patients still develop recurrence or metastasis after treatment. What else can we do?

The tumor microenvironment (TME) of NPC is heavily infiltrated by non-malignant lymphocytes. What's more, PD-L1 is expressed on tumor cells in 83%-92% of patients. These features make NPC an ideal candidate for immune checkpoint blockade. Over the last few years, several phase 3 trials, including our own CONTINUUM and DIPPER trials, have demonstrated that adding PD-1 inhibitors to standard chemoradiotherapy significantly improves event-free survival in locally advanced NPC. However, not all patients benefit equally– about 15% still relapse and 4%-10% experience severe immune-related adverse events. The urgent question is, can we identify patients that will actually benefit from adding immunotherapy? Unfortunately, classic biomarkers like PD-1, PD-L1 expression don't work well in NPC, we need something new. 

Our team previously classified NPC into 3 immune subtypes based on MTME features, but those couldn't predict long-term survival benefit in the context of immunotherapy. By doing differential analysis, we found that relapsed tumors after immunotherapy showed metabolic dysregulation, so we decided to explore metabolic heterogeneity. We collected 407 tumor samples from CONTINUUM and DIPPER studies and performed RNA sequencing. Using unsupervised clustering of metabolic genes, we identified 3 optimal subtypes, which record MS1, 2, and 3. These subtypes showed distinct metabolic features: MS1 had activated an amino acid catabolism, MS2 showed specific activity in polyunsaturated fatty acid metabolism, and MS3 displayed broad activation of amino acid, fatty acid, nucleotide, and energy metabolisms. The classification was robust across different clustering algorithms like K-means and NMF and it was validated in independent cohorts.

We also dug into the underlying biology. MS1 tumors showed activated NF-kB and interferon pathways and were enriched for both cytotoxic and immunosuppressive immune cells with end glycosylation pathways linked to CD8+ T-cell exhaustion. MS2 tumors were enriched in mast cells and metabolic pathways involved in bioactive lipid mediator biosynthesis. MS3 tumors showed increased activity in TGF-beta, EGFR, EMT, MIK, and hypoxia pathways, coupled with DNA methylation-driven transcriptional repression that silences immune-related genes leading to an immune desert phenotype.

When we looked at prognosis, MS2 had the best outcomes. In the chemoradiation therapy alone arm, MS1 and MS3 had similarly poor outcomes. But in the anti-PD-1 plus chemoradiotherapy arm, MS1 patients showed improved 3-year event-free survival from 69.6% to 90.2%. For MS3 patients, the 3-year event-free survival remained at about 75%, regardless of whether immunotherapy was added. In other words, only MS1 patients derived a long-term survival benefit from adding immunotherapy.

What does this mean for clinical practice? For an MS1 patient, we have strong evidence to add PD-1 blockade. For an MS2 patient, given the excellent prognosis with chemoradiotherapy alone, we can confidently spare them from unnecessary immunotherapy and potential toxicity. For an MS3 patient, standard anti-PD-1 adds no meaningful benefit. The immune desert metabolically hyperactive tumor suggests that we need combination strategies such as epigenetic drugs or antiangiogenic agents to rewire the environment.

The good news is that we've already developed a machine learning-based classifier using a small set of genes. It can assign a patient to 1 of the 3 metabolic subtypes from a routine biopsy, and its performance has been validated across both CONTINUUM and DIPPER cohorts. We've also built an online clinical decision platform and via a national pattern in China. With this novel approach, we're now planning a prospective study to test the stratified immunotherapy strategies that we proposed, and I hope that this classifier can eventually be translated into clinical practice to facilitate personalized immunotherapy for NPC patients.

I want to end by thanking our entire team. Special thanks to Dr Ye-Lin Liang, Dr Ya-Lan Tao, and all the co-authors. Our senior investigators Dr Jun Ma, Dr Ying Sun provided invaluable guidance. Of course, we are grateful to the patients who participated in these trials, without them, none of these could be possible.

Source:

Huang SW, Liang YL, Tao YL, et al. Development of a classifier for metabolic subtypes of nasopharyngeal carcinoma to guide personalized immunotherapy strategies: Biomarker analysis of the phase III CONTINUUM and DIPPER Trials. J Clin Oncol. Published online: March 6, 2026. doi:10.1200/jco-25-02111