Recent advances in next-generation sequencing (NGS) techniques enable the quantitation of circulating tumor DNA (ctDNA) encoding the clonal rearranged V(D)J immunoglobulin (IG) receptor gene sequence. The aims of this study were 1) To evaluate the clonal heterogeneity of follicular lymphoma (FL) and its variations between tumor and plasma at diagnosis in patients (pts) included in PRIMA trial, and 2) To assess the prognostic value of the level of ctDNA at diagnosis on PFS (median follow up of 6 years).


Using the NGS-based immunosequencing method (Adaptive Biotechnologies, South San Francisco, CA), the lymphoma clonotype was first established in the tumor biopsy using locus-specific primer sets for IGH-V, IGH-D and IGK rearrangements. Clonotypes regarded as originating from the lymphoma clone were present at a frequency greater than 5%. The lymphoma-derived sequences identified in the biopsy sample were then used as targets to assess the presence of ctDNA in plasma samples. The quantity and frequency of the ctDNA in plasma is calculated relative to the total number of reads in the sample and defined as lymphoma clonotype molecules per million diploid genomes. For 34 FL pts from the PRIMA trial, both tumor biopsy and plasma samples at diagnosis were available.


One tumor clonotype or more could be detected in 29 pts (85%) in the tumor diagnostic sample (DNA isolated from fresh frozen tissue). Indeed, with the IGH-V assay, between 1 to 3 different clonotypes were identified in 23 pts. With the IGH-D assay, 1 clonotype was detected in 2 pts. With the IGK assay, 1 to 2 different clonotypes were detected in 17 pts. The tumor clonotypes were detected in diagnostic plasma samples in 25 out of the 29 pts (83%). Moreover, in 14 pts we detected clones in the plasma samples that are related (with point mutations in the V(D)J sequence) to the highest frequency index clone identified in the tumor sample. In half of the cases, we observed a different repartition of the subclonal populations between tumor and plasma samples at diagnosis: for these cases the highest frequency clone in the tumor was not the highest frequency clone in the plasma.

Among the 24 pts with an IGH clonotype (IGH-V or IGH-D assay), ctDNA could be detected in 19 cases. We could not find a correlation between peripheral blood dissemination and the presence of a stereotyped sequence or an N-glycosylation motif within the CDR3 region of IGH clonotypes. Interestingly, in one case, two different related clones were detected at high frequency in tumor sample and the putative parental clone, not detected in the tumor sample, was detected in plasma sample as the highest frequency clone.

The level of ctDNA ranged from 0 to 345,000/million diploid genomes in the 29 plasma samples (median 39,720). The presence of ctDNA was correlated with bone marrow involvement (p=.005), but not with FLIPI score, LDH level, anemia, bulky disease, presence of circulating lymphoma cells (detected by morphology or flow examination), Ann Arbor stage or β2microglobulin level. The absolute level of ctDNA in diagnostic samples did not correlate with any of the clinical characteristics. We assessed the prognostic value of the level of ctDNA at diagnosis. Pts (14/29) with higher levels of ctDNA at diagnosis (>40,000/million diploid genomes) had shorter PFS than patients with lower levels of ctDNA (median 15.7 m vs NR, p=.027, Fig 1). In an exploratory multivariate Cox regression model including FLIPI score, bulky disease and β2-microglobuline level, high ctDNA level was the only factor significantly associated with a worse PFS (HR 4.2, p=.039). In the observation arm, pts with high ctDNA levels had a median PFS of only 11.6 months vs NR (p=.01). Whereas in the maintenance arm, the prognostic value of elevated ctDNA levels was erased (p=.6).


Using the NGS-based immunosequencing method, we were able to show a variation in clonal heterogeneity between tumor biopsy and plasma samples obtained at diagnosis. The poor prognostic value of elevated ctDNA levels in plasma at diagnosis appeared higher than other clinical parameters in a multivariate cox regression model. These findings need to be validated in larger cohorts.

Figure 1.

PFS according to the level of circulating tumor DNA

Figure 1.

PFS according to the level of circulating tumor DNA


Carlton:Adaptive Biotechnologies Corp.: Employment, Equity Ownership. Faham:adaptive biotech: Employment, Other: stockholders. Salles:Roche: Honoraria, Research Funding.

Author notes


Asterisk with author names denotes non-ASH members.