Gene expression analysis for assessment of lymph node metastasis in head and neck squamous cell carcinoma

Abstract

Head and neck squamous cell carcinoma (HNSCC) consists of a group of neoplasms that arise in the mucosal lining of the upper aero digestive tract. Treatment of HNSCC depends largely on assessment whether the patient has already developed lymph node metastases. However, due to difficulties in detecting cervical lymph node metastases reliably, many patients now receive inappropriate treatment which is accompanied by severe discomfort and complications such as shoulder disability. To improve the diagnosis of lymph node metastasis and to protect a significant number of patients from unnecessary neck surgery, current diagnostic methods need to be improved or new technologies need to be implemented. The primary goal of the work described in this thesis was to investigate the use of DNA microarray gene expression profiling of the primary tumor for assessment of lymph node metastasis in patients with oral cavity or oropharynx cancer. To identify a gene expression signature capable of predicting the presence of lymph node metastasis, a set of primary HNSCC tumor samples was analyzed for genome-wide gene expression patterns using DNA microarrays. Based on a supervised classification procedure, a set of 102 genes, was identified that showed optimal predictive accuracy. The accuracy was validated on independent samples and outperformed the current clinical diagnosis of lymph node biopsy. Most importantly for clinical relevance, no false negative predictions were made in this independent test of the signature. This predictive gene set indicates that treatment of HNSCC patients can substantially be improved by applying the DNA microarray predictor. Comparison of whole genome expression profiles from primary tumors with their corresponding lymph node metastasis indicates that gene expression patterns in lymph node metastases were most similar to the primary tumor from which they originate. Remarkably, both general and metastasis specific primary tumor gene expression patterns are maintained upon spread to the lymph nodes. Apparently, disseminated primary tumor cells do not need to undergo additional developmental changes for survival and proliferation in metastatic sites. This supports the theory that metastatic properties are acquired early during tumorigenesis and sustained throughout cancer progression. Despite the fact that many of the signature genes are expressed in the stroma, primary tumor samples with lower amount of tumor cells were more difficult to classify. Next, we showed that the decreased predictive accuracy for samples with low tumor content was caused by a skewed contribution of the predictor genes. Metastasis of HNSCC primary tumor is predominantly characterized by down-regulation of tumor cell specific genes and a concomitant exclusive upregulation of stromal genes. Correction of the skewed distribution resulted in an increased predictive accuracy, especially for low tumor percentage samples, and thus leads to an increase in samples which can be included in such analyses. Finally, in this thesis some important issues are discussed that need to be considered for optimal design of DNA microarray expression profiling studies in general and for implementation of diagnostic gene expression signatures in the clinic in particular.