Abstract
Introduction: Large B-cell lymphomas (LBCL) exhibit intra-patient biologic heterogeneity. Single-site biopsies limit comprehensive assessment of this complexity, which circulating tumor DNA (ctDNA) overcomes by detecting mutations from distinct anatomical sites. We have previously shown anatomical genomic heterogeneity can be quantified by integrating baseline ctDNA and tissue profiles (Goldstein, ASH 2024). Though aberrant somatic hypermutation (aSHM) drives lymphomagenesis, its contribution to anatomical heterogeneity, clonal evolution, and outcomes is poorly understood. We validate the clinical value of anatomical genomic heterogeneity in diverse aggressive lymphomas, delineate its relation to aSHM, and then propose a DLBCL clonal evolution model.
Methods: We studied 588 specimens from 196 patients across 3 cohorts with comprehensive genotyping of baseline plasma and matched tumor tissues. The Discovery Cohort included 66 LBCL patients from a multinational cohort receiving front-line chemoimmunotherapy with pre-treatment tumor, germline and plasma genotyping by CAPP-Seq (Kurtz, JCO 2018). We developed the Metric Of Spatial Anatomic Intra-tumoral genomic Complexity (MOSAIC) as the root mean square error of a linear regression model of mutant allele frequencies (log[VAF]) in ctDNA versus those shared in tissue.
After defining a MOSAIC threshold in our Discovery Cohort, two validation cohorts with pre-treatment tumor, germline and plasma samples and treated with front-line chemoimmunotherapy were used: 97 LBCL patients from the HOVON-902 study profiled by PhasED-seq (Wang, ASCO 2025) and 33 MCL patients from the LyMa trial profiled by CAPP-Seq (Tessoulin, 18-ICML).
To elucidate clonal evolution, we explored the molecular associations and relationship of aberrant somatic hypermutation (aSHM) with anatomic heterogeneity and outcomes in LBCL. We defined aSHM rate as the fraction of ctDNA mutations occurring within canonical aSHM targets (Schmitz, NEJM 2017). We measured clonal architecture (sciclone) and molecular subtypes (LymphGen) and performed survival analysis (Kaplan-Meier; Cox models).
Results: In our Discovery Cohort, 100% had SNVs unique to plasma, with median 15% of plasma mutations absent in paired tissue. The median MOSAIC was 0.235 (threshold for validation) with range 0.1-0.4. Cases with >1 subclone in plasma had higher MOSAIC (median 0.26 vs 0.19,p=0.002). MOSAIC did not correlate with age, COO, histological subtype, Stage, or IPI. However, high MOSAIC was associated with worse PFS (HR=4.3,p=0.01) and OS (HR=12.7,p=0.02). MOSAIC did not correlate with ctDNA burden and remained prognostic independent of ctDNA levels.
In the MCL LyMa Validation Cohort, median MOSAIC was 0.28 (IQR 0.13-0.39). MOSAIC correlated with ctDNA burden (p=0.05) and LDH (p=0.01), but not with leukemic status, blastoid morphology, or Ki67. High MOSAIC predicted worse PFS (HR 13.6,p=0.01), independent of LDH and ctDNA level.
In the LBCL HOVON-902 Validation Cohort, median MOSAIC was 0.21 (IQR 0.17-0.28) with 38% classified as high. BN2, EZB & A53 tumors by LymphGen had higher MOSAIC (median 0.26), while MCD had lower scores (median 0.18). High MOSAIC was again associated with inferior PFS (HR=2.3,p=0.04) and OS (HR=3.4,p=0.02) and higher end of treatment MRD+ rates (28% vs 12%).
TP53 and P2RY8 mutations were associated with high MOSAIC patients. GCB tumors showed numerically higher aSHM rates, though no significant differences in aSHM rates by histology or stage were found. Lower aSHM rate correlated with higher MOSAIC (r= -0.4,p=0.001) and worse PFS (HR 4.4,p=0.02). Patients with both high MOSAIC and low aSHM rate (38%) had 5-year PFS 43% (HR 5.6,p=0.001) and OS 57% (HR 9.8,p=0.003), vs 87 and 93% respectively. In 80%of patients, median plasma VAF for aSHM exceeded non-aSHM mutations, indicating aSHM occurs early in lymphomagenesis. High MOSAIC patients had higher ratios of median aSHM to non-aSHM VAFs, consistent with ongoing subclonal evolution via non-aSHM mutations.
Conclusions: We describe and validate a novel biomarker measuring anatomical genomic heterogeneity, integrating plasma and tissue profiles, that independently predicts prognosis for front-line chemoimmunotherapy. We show that aSHM aberrations occur early in clonal evolution and likely initiate lymphomagenesis. Low aSHM rates and high anatomical heterogeneity in B-cell lymphomas define a high-risk subgroup prone to treatment resistance via anatomical clonal divergence.
