Background: Follicular lymphoma (FL) is usually an indolent malignant B-cell lymphoma with median survival now approaching 20 years in the rituximab treatment era. However, at an annual rate of 2-3%, FL patients experience transformation to Diffuse Large B-cell Lymphoma (DLBCL) with inferior survival. Early identification of patients with high risk of transformation would open the possibility of tailored treatment. We recently developed a statistical method inspired by the Copy Number and Expression in Cancer (CONEXIC) algorithm (Akavia et al, Cell 2010) to identify putative drivers in cancer by integration of DNA copy number, mRNA gene expression and prior knowledge of protein network interactions in FL (Brodtkorb et al, Blood 2014). This method identified six copy number driven genes that are upstream regulators of NF-κB, and their corresponding gene signature scores were predictive of transformation in the pre-rituximab treatment era. Here, we further developed this analysis to rituximab treated FL patients.

Material and Methods: We performed a network-based integrative analysis of 168 FL biopsies from 127 patients with long clinical follow-up and transformation status from two cohorts. Cohort I included 44 patients diagnosed with FL and who were mainly treated prior to the introduction of rituximab. Cohort I was selected for transformation cases in particular, and 61% of these patients transformed during a median follow-up of 88 months (range 10-294) (Eide et al, Blood, 2010; Brodtkorb et al, Blood, 2014). Cohort II consisted of 83 patients diagnosed with FL and included in two prospective clinical trials with rituximab with or without interferon α in first line (Kimby et al., Leukemia Lymphoma, 2008 and 2015). During follow-up with median observation time of 129 months (range 6-312), 22 of the patients in Cohort II (27%) experienced transformation to more aggressive disease. Both cohorts had available DNA copy number profiles (array CGH, Affymetrix SNP6.0) and gene expression profiles (Affymetrix Gene Chip HGU 133 Plus 2.0) from FL biopsies: 75 in Cohort I, and 93 in Cohort II. In Cohort II, 85 biopsies were sampled before start of rituximab treatment, and eight were sampled at relapse, before new treatment was administered. Gene signatures were first analyzed by calculating the gene signature scores as described in Brodtkorb et al. (Blood, 2014) and testing for association to transformation. We subsequently extended the analysis by calculating the principal components of the gene signatures. All biopsies were included in the analyzes, also when the same patient had several biopsies. Survival analysis (Kaplan-Meier plots) was performed in Cohort II, keeping the earliest high-score biopsy for cases with serial biopsies, and using a Log-rank test to assess the significance.

Results: Integrative analysis of copy number and gene expression data from the 168 biopsies revealed a consistent pattern of copy number driven gene expression alterations in the two independent treatment cohorts. A set of NF-κB related gene expression scores previously found to predict transformation in Cohort I also predicted transformation in patients treated with rituximab (Cohort II). We found a shift in expression towards DLBCL for FL biopsies from patients with subsequent transformation, and this shift could be detected years before transformation. Combining a gene expression score with clinical parameters, we were able to identify a subgroup of patients (16%) with short time to transformation (P=0.000003), as well as short time to new treatment for relapse after rituximab without chemotherapy in first line (P=0.00434). The median time to new treatment was 22.7 months (95% CI: 14.3-42.6) in the high-risk group versus 70.9 months (95% CI: 41.5-95.0) for the rest of the patients.

Conclusion: Integrative analysis identified molecular markers for transformation risk and a gene expression score predicted transformation in FL-patients treated with rituximab in first line.Combination of an expression marker with clinical features improved the early identification of FL patients with high risk of transformation and risk of relapse.


Kimby: Roche: Honoraria, Membership on an entity's Board of Directors or advisory committees, Research Funding; Abbvie: Membership on an entity's Board of Directors or advisory committees; Gilead: Honoraria; Janssen: Honoraria, Membership on an entity's Board of Directors or advisory committees; Celgene: Research Funding; Pfizer: Research Funding. Holte: Oslo University Hospital: Employment; Nordic Nanovector: Consultancy; Novartis Pharmaceuticals Corporation: Membership on an entity's Board of Directors or advisory committees.

Author notes


Asterisk with author names denotes non-ASH members.