Gene Expression Profiling Can Predict Responsiveness to Chemotherapy in Patients with Transformation of Follicular Lymphoma.

Background. Transformation to diffuse large cell lymphoma (DLCL) is a frequent event in patients with follicular lymphoma (FL), occurring in approximately 30–50% of patients. Upon transformation the prognosis of these patients is dismal, with a median overall survival of 1–2 years. However, a subset of patients does respond to aggressive chemotherapy regimens, including autologous stem cell transplantation and can achieve long-term disease-free survival. Currently, there are no prognostic factors reliably predicting response to treatment. We used gene expression profiling to identify a profile which can predict responsiveness to chemotherapy and long-term survival at transformation. Patients and methods. From the pathology archives of the 4 participating institutions, 46 cases with transformation of FL were identified for whom frozen tissue was available. 11 cases were excluded from the analysis: 4 because of insufficient clinical information and 7 because the patients were not treated with aggressive chemotherapy regimens (defined as containing at least CHOP-like chemotherapy). RNA was extracted from frozen tissue. All samples were hybridized to cDNA microarray slides prepared at the Central Microarray Facility at the Netherlands Cancer Institute. The arrays contain 18336 biological transcripts as well as 864 control probes. Samples were cohybridized with a tonsil reference. Data analysis was performed in the statistical package R as well as with BrB array tools. Results. Of the 35 patients, 62% was male. The median interval between the initial diagnosis of FL and transformation was 27 months (range 0–252; 4 patients had transformed FL at initial diagnosis). Median age at transformation was 52 years (range 32–78). Using hierarchical clustering, no clear separation of the cases was obtained. Of the 35 cases, 13 reached a CR or Cru and had an overall survival of >24 months (range 37–134+, median 60 months). 6 patients achieved a partial remission and had an overall survival of 12–24 months, and 16 patients had stable or progressive disease and died within 13 months (range 3–13, median 5.5 months). Using supervised analysis on patients grouped based on survival, a classifier with an accuracy of up to 88% accuracy could be reached in an LOOCV procedure. An analysis of the most differentially expressed genes related to survival identified mostly genes involved in cell cycle, metabolism and immune related genes. Conclusions. Using gene expression profiling a classifier can be constructed that can separate transformed FL with a long survival from transformed FL with a short survival. This classifier can help to identify patients who benefit from intensive chemotherapy. Functional annotation of the differentially expressed genes identified mostly genes involved in cell cycle control and metabolism, and immune related genes. We plan to validate this approach using immunohistochemistry.