PURPOSE:
To find molecular markers from expression profiling data to predict recurrence of laryngeal cancer after radiotherapy.
EXPERIMENTAL DESIGN:
We generated gene expression data on pre-treatment biopsies from 52 larynx cancer patients. Patients developing a local recurrence were matched for T-stage, subsite, treatment, gender and age with non-recurrence patients. Candidate genes were then tested by immunohistochemistry on tumor material from a second series of 76 patients. Both series comprised early stage cancer treated with radiotherapy alone. Finally, gene expression data of eight larynx cancer cell lines with known radiosensitivity were analyzed.
RESULTS:
Nineteen patients with a local recurrence were matched with 33 controls. Gene sets for hypoxia, proliferation and intrinsic radiosensitivity did not correlate with recurrence, whereas expression of the putative stem cell marker CD44 did. In a supervised analysis, probes for all three splice variants of CD44 on the array appeared in the top 10 most significantly correlated with local recurrence. Immunohistochemical analysis of CD44 expression on the independent validation series confirmed CD44’s predictive potential. In 8 larynx cancer cell lines, CD44 gene expression did not correlate with intrinsic radiosensitivity although it did correlate significantly with plating efficiency, consistent with a relationship with stem cell content.
CONCLUSIONS:
CD44 was the only biological factor tested which significantly correlated with response to radiotherapy in early stage larynx cancer patients, both at the mRNA and protein levels. Further studies are needed to confirm this and to assess how general these findings are for other head and neck tumor stages and sites.
TRANSLATIONAL RELEVANCE:
Treatment choice for larynx cancer is based on clinical factors such as T-stage, but these are imprecise indicators of response. Having robust methods to predict outcome of a particular therapy would be extremely valuable, allowing a more rational treatment choice which should lead to greater tumor cell kill and also spare patients from toxic and ineffective therapies. Such predictors should include biological factors as well as clinical factors, given the heterogeneity in tumor biology even for patients presenting with similar sites and stages. The present study employed gene expression profiling in a series of larynx cancers and validated the result in a second similar series using immunohistochemistry. The principle predictor for outcome after radiotherapy was CD44, a putative stem cell marker. In addition, this study sheds light on potential mechanisms of radioresistance, which could lead to the design of targeted drugs for combining with radiation.