Somatic Mutations Detected in Plasma Cell-Free DNA By Targeted Sequencing: Assessment of Liquid Biopsy in Primary Central Nervous System Lymphoma

Background

Primary Central Nervous System Lymphoma (PCNSL) are rare and aggressive primary brain tumors. Histological diagnosis can be difficult at initial stages or at relapse due to deep brain structure involvement. Finding a minimally invasive biomarker aiding the diagnosis remains an unsolved question. Plasma Cell-Free DNA (cfDNA) seems to have shown its diagnostic and prognostic value in nodal Diffuse Large B Cell Lymphomas (DLBCL) [Roschewski et al, Lancet Oncology, 2015; Kurtz et al, Blood, 2015]. Our main objective was to demonstrate that targeted sequencing of cfDNA in plasma at time of diagnosis could identify PCNSL somatic mutations.

Methods

30 immuno-competent patients suffering from newly diagnosed PCNSL, without any extranevraxic lesions, were enrolled from 2008 to 2014. Tumor tissues and plasma samples were collected at the time of diagnosis and frozen until use. High throughput sequencing was performed on primitive tumors using a panel of 34 genes relevant to lymphomagenesis, as previously reported [Dubois et al, Oncotarget, 2015; Bohers et al, Haematologica, 2015]. We next performed patient-specific targeted sequencing of identified somatic mutations in cfDNA. The detection sensitivity threshold was set at 1% for all SNVs, except for MYD88L265P, which was set at 0.1%. The primary endpoint was the proportion of patients having at least one somatic mutation found in the plasma.

Results

Among 24 available plasmas, 15 patients (63%) had at least one detected somatic mutation in cfDNA. All plasmas had detectable cfDNA (mean concentration 1.6 ng/µL). No correlation was found between tumor volume and cfDNA concentration (R squared coefficient 0.01). Regarding the whole sequenced cohort (n=30) 21 (70%) were classified as nonGC subtype, 8 (27%) as GC subtype and 1 patient (3%) as unclassifiable, according to the Hans algorithm. 29 tumors had at least one somatic mutation, mainly nonsynonymous single nucleotide variants (SNV). The NF-kB pathway was the most affected by mutations: MYD88 (n=23, 77%), PIM1 (n=11, 37%), TNFAIP3 (n=6, 20%), IRF4 (n=3, 10%), CARD11 (n=3, 10%) and PRDM1 (n=3, 10%). Among the 23 tumors harboring a MYD88 mutation, the L265P variant was the most frequent (20 patients, 67%); mean tumor variant allele frequency was 46% [min 8%, max 91%]. One tumor harbored a single MYD88 L265P mutation with no other detectable abnormality. Among patients with both available plasma and a somatic MYD88 L265P mutation in the tumor, 15 patients (88%) had an identifiable L265P variant in cfDNA, with a mean variant allele frequency of 4% [min 0.1%, max 28%]. PIM1 and TNFAIP3 SNVs were also detected in cfDNA for respectively two and one patient.

The second most affected pathway was the apoptotic pathway: genes affected by mutations included GNA13 (n=7, 23%), TP53 (n=2, 7%), MYC (n=1, 3%), CDKN2A (n=2, 7%) and BCL2 (n=1, 3%). The B Cell Receptor (BCR) pathway was also affected, mainly due to mutations targeting CD79B (n=10, 33%) and ITPKB (n=3, 10%) mutations. 8 tumors (27%) harbored a dual alteration affecting MYD88 and CD79B. One tumor of the GC subtype had one EZH2 SNV (Y646H), but the mutation was not found in cfDNA.

There was no significant difference in overall survival (OS) between patients with and without mutations detected in cfDNA: mean OS 27 months versus 18 months (HR 0.6; IC95% [0.2 – 1.6], p value 0.3).

Tumor gene copy number variations were detected in 29/30 patients with either homo or heterozygous deletions or copy gains. CDKN2A/2B deletions were detected in 23 cases (77%).

Conclusion

To our knowledge this is the first study that provides evidence that somatic mutations can be detected in cfDNA in patients suffering from PCNSL, therefore constituting a minimally invasive tool helping for diagnosis. Further studies are now required to improve prognosis and predictive values of this new promising procedure for PCNSL patient care.