Joint Analyses of Serum Proteins and Circulating Tumor DNA Improves Risk Assessment and Subtype Stratification in Large B-Cell Lymphomas

Abstract

Introduction Inadequate dismantling of clinical and biological heterogeneity is an obstacle for improving outcome in large B-cell lymphoma (LBCL). Circulating tumor DNA (ctDNA)-based stratification holds great potential in resolving these challenges. However, the assessment is limited to genetic content, and excludes host response from the estimates. Previously, we have shown that an inflammatory signature identified from the pretreatment serum proteins translates to poor survival, and multiple serum proteins correlate with LBCL genetic subtypes. To further improve risk stratification and molecular characterization of LBCL from the liquid biopsy, we co-analyzed serum protein and ctDNA levels in LBCL patients treated in two Nordic Lymphoma Group (NLG) trials. Materials and methods Our discovery cohort consisted of 122 patients less than 65 years of age with clinically high-risk (age adjusted IPI ≥ 2) LBCL, who were treated in NLG-LBC-06 trial with immunochemotherapy. Plasma and serum samples were collected prior to treatment. Cell free DNA was extracted from plasma and sequenced with in-house adapted duplex sequencing platform using custom targeted panel of 347 lymphoma driver genes and immunoglobulin loci. An inflammation score was composed as log transformed mean of the most pertinent inflammatory serum proteins detected in the inflammatory signature (IL10, IL2RA, CXCL9, IL18, GZMB, PD-L1 and IFNG), and their levels were measured using antibody-based Luminex assay (pg/ml, R&D Systems). In addition, CCL17 and TNFRSF13B levels were analyzed to uncover their associations with genetic subtypes. An independent cohort of 91 patients treated in our previous NLG-LBC-05 trial was used for validation. Results Combined analysis of serum proteins and ctDNA identified a group of patients (n=12, 10%) with high serum inflammation scores and ctDNA burden (≥ 3.75 hGE/ml), which translated to poor survival compared to the patients with low concentrations of either inflammatory proteins or ctDNA, or both (progression free survival, PFS, hazard ratio 7.33, 95%-CI 2.59-20.79, P < 0.001, Figure 1A). This finding was confirmed in the independent validation cohort. In both cohorts, the combined analysis showed higher specificity in survival prediction as compared to the ctDNA-based assessment alone, indicating that inflammatory serum markers alongside with ctDNA can identify those patients who, despite of having high ctDNA burden, have excellent survival. Beyond inflammation, the levels of serum TNFRSF13B and CCL17 correlated with mutations associating with non-germinal center diffuse large B-cell lymphoma (non-GCB DLBCL)/MCD subtype and primary mediastinal B-cell lymphoma/grey-zone lymphoma (PMBCL/GZL), respectively. For example, serum TNFRSF13B levels correlated positively with mutations in MCD genes, such as CD79B, SETD1B, PIM2 and PRDM1, and negatively with mutations in EZB genes EZH2 and CREBBP in both cohorts (P < 0.05, Figure 1B). Supporting our earlier findings, high CCL17 levels revealed a group of patients diagnosed with DLBCL not otherwise specified or high-grade lymphoma but with a mutational landscape resembling PMBCL/GZL, including mutations in TNFAIP3, ITPKB, B2M, BTG1 and STAT6 (P < 0.05). Conclusion Our findings demonstrate that co-analysis of inflammatory serum markers along with the ctDNA improves ctDNA-based survival prediction. Moreover, analysis of serum proteins refines molecular classification. Altogether, joint analysis of serum proteins and ctDNA provides new tools to dismantle LBCL heterogeneity.