Combined Genotype Analysis of Tumor and Cell-Free DNA By Ultra-Deep Targeted Sequencing: Correlation with PET-Scan Imaging in a Prospective Cohort of Diffuse Large B-Cell Lymphoma Patients

The emerging concept of liquid biopsy in diffuse large B-cell lymphoma (DLBCL) is based on the release of Cell-free DNA (cfDNA) into the blood by tumor cells undergoing apoptosis or necrosis that can be used for the identification of mutations (Bohers et al., Haematologica 2015 ; Rossi et al., Blood 2017). cfDNA theoretically can give therefore information regarding potentially actionable mutations, clonal evolution, and genetic mechanisms of resistance. To determine the clinical relevance of cfDNA, we assessed prospectively by Ultra-deep targeted next generation sequencing (NGS) the basal DLBCL genetic pattern, its modification upon treatment using plasma cfDNA analysis and its correlation with PET scan imaging (clinical.gov number: NCT02339805).

Methods

Thirty consecutive untreated DLBCL or primary mediastinal B-cell lymphoma (PMBL) were enrolled and followed after rituximab-cyclophosphamide-doxorubicin-vincristine-prednisone (R-CHOP) or R-CHOP-like chemotherapies. Mutational status of a 34-genes lymphopanel (CIITA, B2M, TNFRSF14, CD58, NOTCH1, NOTCH2, MFHAS1, XPO1, MYC, CDKN2A/B, FOXO1, TP53, GNA13, BCL2, TNFAIP3, MYD88, PIM1, CARD11, IRF4, PRDM1, EZH2, KMT2D, EP300, MEF2B, CREBBP, BRAF, CD79A/B, ITPKB, TCF3, ID3, SOCS1 and STAT6) was assessed in both tumor (genomic DNA, gDNA) and cfDNA at the time of diagnosis using AmpliSeq®/PGM® technologies (Dubois et al., Clinical Cancer Research 2016) . Longitudinal analysis of plasma samples collected at mid-treatment (C4) and at the end of treatment or at relapse/progression was performed and correlated with PET scan characteristics obtained at baseline (Metabolic tumor volume, MTV), mid and end-treatment (Delta SUV and Deauville score). To increase the NGS sensitivity and for cost-effectiveness reasons, only mutations detected at baseline were monitored during the follow-up. Variant analysis was performed using an in-house generated bioinformatics pipeline.

Results

Patient clinical features were as follow: IPI: 0-1, n =8; IPI: 2-3, n=14; IPI: 4-5, n =8; median age was 64 y; bone marrow involvement: 0/25 patients; elevated LDH: 12/30; Fifteen patients were sub-classified as GCB, 9 as ABC/non GCB and 3 as PMBL. cfDNA was available for the 30 patients at the time of diagnosis. Mean plasma concentration was 74 ng/mL (14.6 - 433) and sequencing was carried out with a mean depth of 3750x. cfDNA mutations were identified in 20 patients (67%) and were consistent with the pattern usually observed in the different subtypes of DLBCL. gDNA mutations from tissue biopsies (11 frozen/14 FFPE) were found in 22/25 (88%) cases. In 3/22 cases for whom gDNA analysis failed, an informative mutational pattern was finally successfully obtained in matched cfDNA. Variant allelic frequency (VAF) was lower in cfDNA (mean 15%) compared to tumor DNA (mean 33%). Finally, among the 22 tumor/plasma matched samples, 8 patients had no detectable tumoral cfDNA, whereas 4 patients displayed additional mutations in plasma compared to tumor. At the time of diagnosis, a high [cfDNA] correlated with the presence of tumoral circulating DNA and cfDNA mutations were observed in all patients with elevated LDH. MTV correlated with the highest concentrations of cfDNA and VAF rate. Moreover, 3 / 4 patients presenting additional mutations in cfDNA, as compared to gDNA, had a high MTV (> 2000cm³), suggesting that cfDNA mutations reflect more accurately tumor heterogeneity than gDNA analysis. After R-CHOP/ R-CHOP-like regimens, a rapid clearance of cfDNA mutations was observed in 13/17 cases. Conversely, among the 4/17 patients whose basal DLBCL mutations did not disappear at mid-treatment, 3 were in partial response according to Deauville score, including 2 with a delta SUVmax < 70%. One patient in partial remission had progressive disease 6 months after the end of the treatment and still harboured tumoral cfDNA, as shown by a persistent detectable TP53 mutation.

Conclusion

Our study demonstrates that cfDNA genotyping of DLBCL is an accurate genotyping tool and represents a real-time and non-invasive approach to tracking treatment-resistant clones. It highlights the major interest of the liquid biopsy in the context of bulky tumors where cfDNA is more representative than gDNA by capturing the entire tumoral mutation profile. cfDNA analysis represents therefore a complementary approach to TEP scan imaging at baseline and during follow-up for the management of DLBCL.