Abstract

Introduction. Accessible and real-time genotyping for diagnostic, prognostic or treatment purposes is increasingly impelling in diffuse large B-cell lymphoma (DLBCL). Since DLBCL lacks a leukemic phase, tumor genotyping has so far relied on the analysis of the diagnostic tissue biopsy. Cell-free DNA (cfDNA) is shed into the blood by tumor cells undergoing apoptosis and can be used as source of tumor DNA for the identification of cancer-gene somatic mutations. Accessing the blood has obvious sampling advantages in the monitoring of mutations in real-time. Also, cfDNA is representative of the entire tumor heterogeneity, thus allowing to identify mutations from tumor cells residing in non-biopsied sites. Here we aimed at tracking the DLBCL genetic profile using plasma cfDNA.

Methods. The study was based on 26 consecutive DLBCL patients (age >65=15, male:female ratio=11:15) presenting in different Ann Arbor stages (III-IV=13) and age-adjusted IPI risk scores (2-3=13), and provided with cfDNA from plasma collected at diagnosis, during R-CHOP course, at the end of treatment and at progression. Paired normal genomic DNA from granulocytes was also collected for comparative purposes to filter out polymorphisms. A targeted resequencing panel including the coding exons and splice sites of 59 genes (207 kb) that are recurrently mutated in mature B-cell tumors was specifically designed to allow the recovery of at least one mutation in >90% of DLBCL. Ultra-deep next-generation-sequencing (NGS) of the gene panel was performed on MiSeq (Illumina) (coverage >2000x in >80% of the target) using a SeqCap library preparation strategy (NimbleGen). The somatic function of VarScan2 was used to call non-synonymous somatic mutations, and a stringent bioinformatic pipeline was developed and applied to filter out sequencing errors. The study cohort was divided in a training set of 17 patients provided with paired tumor DNA from the diagnostic tissue biopsy that was used for the set up of the ultra-deep NGS strategy, and an extension set of 9 patients lacking the tissue biopsy.

Results. A total of 76 cfDNA samples (26 pretreatment, 28 during treatment, 18 at the end of treatment, and 4 at time of progression) were evaluated. Pretreatment cfDNA genotyping disclosed somatic mutations of heterogeneous abundance (median mutated molecules/ml of plasma: 3168, range 1.73-6.5x104) in known DLBCL-associated genes, including MLL2 (33%), TP53 (25%), CREBBP and TNFAIP3 (21%), EZH2, TBL1XR1, PIM1 (17%), B2M, BCL2, CARD11, CCND3, FBXW7 and STAT6 (13%) (Fig. 1A). The results of genotyping on cfDNA from plasma and of genomic DNA from tumor cells of the diagnostic biopsy (gold standard) were compared to derive the diagnostic accuracy of cfDNA genotyping (Fig. 1B). Genotyping of the paired plasma cfDNA correctly identified 79% of the tumor biopsy mutations. Most of the tumor variants not discovered in the cfDNA had a low representation in the tumor biopsy (median allelic abundance=5.7%; range 0.8-54%). Consistently, ROC analysis showed that cfDNA genotyping had the highest sensitivity (92%) if mutations were represented in >15% of the alleles of the tumor biopsy. Plasma cfDNA genotyping also disclosed a number of additional somatic mutations (~2 per case, range 1-6) that were not detectable in the tissue biopsy, including mutations of clinically relevant genes. Longitudinal analysis of plasma samples under R-CHOP chemotherapy showed a rapid clearance of the DLBCL mutations in the cfDNA already after the first cycle among responding patients (Fig. 1C). Among patients that were resistant to R-CHOP, basal DLBCL mutations did not disappear from cfDNA. In addition, among treatment-resistant patients, new mutations appeared in cfDNA that conceivably marked resistant clones selected during the clonal evolution process taking place under the pressure of treatment (Fig 1D).

Conclusions. Overall, these results provide the proof of principle that cfDNA genotyping of DLBCL: i) is as accurate as genotyping of the diagnostic biopsy to detect somatic mutations of allelic abundance >15% in DLBCL; ii) allows the identification of mutations that are otherwise absent in the tissue biopsy conceivably because restricted to clones that are anatomically distant from the biopsy site; and iii) is a real-time and non-invasive way to track clonal evolution and emergence of treatment resistant clones.

Disclosures

No relevant conflicts of interest to declare.

Author notes

*

Asterisk with author names denotes non-ASH members.