Background: In the absence of tumor tissue, several studies have shown the utility of plasma cell-free DNA (cfDNA) for blood based noninvasive genotyping of diverse human tumors, including aggressive lymphomas. However, it remains unclear whether these same cfDNA advantages also apply to follicular lymphoma (FL), and if they could inform noninvasive minimal residual disease (MRD) monitoring in the context of modern regimens inducing deep FL remissions. Here, we address these questions in previously untreated patients with FL receiving standard first-line (1L) therapy, using Phased Variant Enrichment and Detection Sequencing (PhasED-Seq, Kurtz et al 2021, Nat Biotech) to monitor ctDNA and MRD. Methods: We profiled 298 samples from 61 patients with FL, including archival formalin fixed paraffin embedded (FFPE) tumor tissues, and serial blood samples before, during, and following 1L therapy. All patients were treatment-naive, with most (75%) presenting with advanced stage FL. Induction regimens primarily comprised Bendamustine/Rituximab (66%) and R-CHOP (26%). At a median follow-up of 23 months from start of 1L therapy, 80% of patients were alive and progression free, while 11% experienced progression within 24 months (POD24). MRD was analyzed by PhasED-Seq (Foresight Diagnostics). Phased variants (PVs) were genotyped either using baseline blood plasma, or alternatively from FFPE tumor tissue; matched leukocytes were used as a source of constitutional DNA to censor germline variants and clonal hematopoiesis of indeterminate potential (CHIP). Baseline PV genotypes from each baseline source were then used to longitudinally assess MRD serially during 1L therapy. Blood timepoints included baseline (pre-treatment), Cycle 2 Day 1 (C2D1), C3D1, C4D1, C5D1, C6D1, at End of Induction (EOI), and serially thereafter. Levels of ctDNA were compared to known FL prognostic factors, including radiographic responses, POD24, and progression-free survival (PFS). Results: PVs were successfully identified from tumor FFPE samples in all patients (61/61), enabling ctDNA MRD assessment by PhasED-Seq. The median number of PVs identified was 530 (IQR 285-1152). Prior to therapy, ctDNA was detectable in 94% of cases (57/61). The median tumor fraction in baseline plasma was 0.28%, significantly lower than pretreatment ctDNA levels in diffuse large B-cell lymphoma (DLBCL) (DLBCL median = 6.1%, Roschewski, ASH 2022, P<0.0001 for FL vs DLBCL). Median pre-treatment ctDNA levels were significantly higher in advanced stage disease (stage I/II, 0.04%, stage III/IV, 0.53%, P=0.006), and correlated with FLIPI risk scores (P=0.04). We next tested whether adequate PVs can be directly identified from plasma without need for tumor tissue to allow MRD surveillance. Despite lower baseline ctDNA levels in FL, a sufficient number of PVs were identified directly in plasma to enable disease monitoring in 75% of cases (45/61). Moreover, ctDNA levels were highly correlated, regardless of the baseline sample used to identify PVs (i.e., tumor versus pretreatment plasma, Spearman rho=0.92, P<0.0001, Fig 1a). Despite high genotypic concordance of tumor vs plasma, we observed significant tumor heterogeneity across compartments with variable rates of concordance between specific mutated genes (Fig 1b). We next assessed the prognostic performance of ctDNA MRD to predict clinical outcomes. ctDNA levels prior to treatment were not significantly prognostic for time to progression. In contrast, in patients with ctDNA MRD assessed after two cycles of therapy (i.e., at C3D1), detection of residual ctDNA was significantly associated with inferior time to progression (log-rank P=0.03 HR = 8.9). Finally, we assessed the feasibility of serial ctDNA surveillance during radiographic remission. In patients with subsequent clinical relapse of disease, we detected reemergence of detectable MRD over two years prior to clinical disease progression. Conclusions: Circulating tumor DNA analysis with PhasED-Seq is feasible in FL, with baseline ctDNA levels significantly lower than those observed in DLBCL. In cases without tissue samples available, plasma DNA can frequently be used as a surrogate for tissue samples. ctDNA detection during treatment is strongly associated with eventual treatment responses, and serial disease surveillance with ctDNA can anticipate clinical relapse.