Circulating Tumor DNA for Primary Mediastinal B-Cell Lymphoma Genotyping: A Prospective Multicenter Study

Introduction: Primary mediastinal B-cell lymphoma (PMBL) is a rare and aggressive B-cell lymphoma. Its diagnosis primarily relies on small core biopsies. Therefore, alternative sources of tumor DNA are needed for genotyping when the tissue biopsy has been exhausted during routine diagnostics.

Methods: In this preliminary analysis, we present the baseline mutational profiling results of patients enrolled in the ongoing prospective CAMIL study (NCT04824950). The study aims to evaluate the diagnostic performance of ctDNA minimal residual disease (MRD) in PMBL. Adult patients newly diagnosed with PMBL and treated with combination immunochemotherapy were prospectively included across 30 LYSA centers between 2021 and 2023. Before administering any anti-lymphoma treatment, except for corticosteroids, blood was collected in Streck® Cell-Free DNA BCT for the isolation of cfDNA and in PAXgene® Blood DNA tubes for the isolation of gDNA of PBMC which was used as germline controls. A Cancer Personalized Profiling by Deep Sequencing (CAPP-seq) ctDNA protocol was used to analyze ctDNA (~380 kb selector, 155 genes, including 30 non-coding regions targeted by aberrant somatic hypermutation [aSHM]; coverage >2000x in >80% of the region of interest). A background error-suppressed approach was used for variant calling (sensitivity: 10^-3). Non-target reads were utilized for the detection of somatic copy number abnormalities (SCNA) and FACTERA was used for the detection of fusions.

Results: The study recruited 87 patients, of which 86 had baseline plasma samples collected before the start of first-line treatment. ctDNA was detected in 84/86 (98%) patients. ctDNA fragmentation patterns showed 53% of patients had predominantly mononucleosomal fragments (median size [range] : 181 [151-208] bp), while 47% had a shift towards submononucleosomal lengths (159 [134-181] bp). The median baseline ctDNA level was 703.7 hGE/mL; the median variant allele frequency was 8.8%; the median number of variants per gene and per sample were 7 and 99, respectively. Among the 133 mutated genes, the top 5 encompassed BCL6 (97.6%, almost exclusively non-coding mutations of the intragenic superenhancer [SE]), SOCS1 (88.1%), IGLL5 (84.5%), CD83 (72.6%, almost exclusively non-coding mutations mapping in intron 2), and B2M (71.4%). The five genes with the highest number of different variants were IGLL5 (n=3631), BCL6 (n=2141), SOCS1 (n=876), CD83 (n=351), and RHOH (n=192), all these genes being targets of aSHM mediated by AID. The top 3 SCNA were focal gain of 9p24.1 (60%), gain of chromosome 9 (23.5%), and focal gain of 2p16.1 (22.4%). Within the genomic space covered by our selector, we identified at breakpoint resolution recurrent genomic translocations involving SOCS1 (n=6), IL21R-IL4R (n=3), and MAP3K14-SPATA32 (n=2). We compared the aSHM load of PMBL with that of classical Hodgkin lymphoma (cHL, n=317) and diffuse large B-cell lymphoma (DLBCL, n=235) cases, all assessed using ctDNA with the same CAPP-seq selector, reusing data from our laboratory's database. The median proportion of variants belonging to the AID-signatures was significantly higher in PMBL (58.3%) compared to cHL (31.6%) and DLBCL (42.5%) (p<0.01). The proportion of PMBL harboring mutations within AID-hypermutated SE hotspots was significantly higher (26.7%) compared to cHL (9.8%) and DLBCL (15.1%) (p<0.05). The AID-hypermutated SE hotspots BCL6, SOCS1 and CD83 were more heavily mutated (95%, 86%, 71%) in PMBL than DLBCL (71%, 17%, 23%) and cHL (66%, 49%, 23%). By performing Nonnegative Matrix Factorization-based clustering, we observed that the majority of PMBL cases (n=47, 55.3%) are grouped with cHL cases (n=75, 31.4%) and a small subgroup of DLBCLs (n=4, 1.7%). This cluster is predominantly characterized by 9p24.1 gains (27% of cases), involving JAK2 and PDL1/PDL2, and other significant genetic alterations such as SOCS1 (80.2%), TNFAIP3 (62.7%), B2M (58.7%) and XPO1 (28.6%) mutations. These genetic features collectively argue for pathogenic pathway associated with immune evasion mechanisms and cytokine signaling, particularly involving the JAK/STAT pathway.

Conclusion: ctDNA analysis by CAPP-seq in PMBL is informative for 98% of patients at baseline and reflected the strong involvement of AID-associated hypermutation in the lymphomagenesis. Each identified variant at diagnosis can serve as a biomarker for monitoring MRD during treatment.