Tracing Founder Mutations in Circulating and Tissue-Resident Follicular Lymphoma Precursors

Follicular lymphoma (FL) is the most common indolent non-Hodgkin lymphoma and is genetically characterized by translocations involving the BCL2 locus. These t(14;18)(q32;q21) translocations are detectable in nearly 90% of tumors driving overexpression of the antiapoptotic protein BCL2 (1, 2). However, as many as 70% of healthy older adults also carry BCL2-translocated cells in blood and tissues (3, 4, 5). Although the presence of clonally expanded t(14;18) at high frequencies predicts an increased risk, the majority of individuals with detectable BCL2 translocations never develop FL (3, 6, 7). Likewise, although overexpression of BCL2 is seen in potential nodal precursor lesions, such as in situ follicular neoplasia (ISFN), fewer than 5% of such patients develop FL (8, 9). BCL2 translocations are therefore critical but not sufficient for FL development. Additional genetic alterations are likely required for lymphomagenesis, as also suggested by murine models (10).

FL samples paired in space (lymph node vs. marrow) or time (diagnosis vs. relapse) often show limited mutational concordance, suggesting that FL evolves and persists from a reservoir of precursor cells bearing early variants (11, 12). The existence of distant circulating FL precursors has been directly confirmed in two studies of clonally related lymphomas in transplant donor–recipient pairs (13, 14). Sequencing studies of FL tumors have identified recurrent and clonally dominant lesions as potential early events (15–20). In addition to the acquisition of novel N-glycosylation sites through somatic mutation of immunoglobulin variable regions (21), an almost ubiquitous feature of FL involves the mutation of chromatin-modifying genes (CMG), including KMT2D, CREBBP, EZH2, and EP300. Potential early lesions have also been identified in ISFN specimens, with alterations in genes recurrently mutated in FL, including BCL2, RUNX1, EBF1, and CREBBP (18, 22). Together, these observations suggest that even clones present years before diagnosis may contain committed cancer precursor cells (CPC). However, there has been no direct systematic characterization of FL precursor mutations in prediagnostic samples, and little is known about the hierarchy and timing of somatic aberrations preceding FL diagnosis.

Similar precursor cells have been identified in myeloid hematopoiesis, with alterations in genes including DNMT3A, TET2, and ASXL1 associated with clonal expansion [clonal hematopoiesis of indeterminate potential (CHIP); ref. 23]. As with BCL2 translocations, such mutations are more frequent with age, portend an increased risk for hematologic malignancies, and are detectable years prior to clinical diagnosis (24). Identification of precursor lesions in a screening context is therefore a promising strategy for risk stratification in myeloid disease (25, 26). However, although these mutations can be traced to early hematopoietic precursors, including stem cells (27), they are less frequently observed in lymphomas, suggesting an alternative genetic repertoire in lymphoid precursors. Detecting such variants presents a technical challenge: A recent large-scale study evaluating myeloid versus lymphoid clonal hematopoiesis uncovered lymphoid mutations primarily in association with subsequent leukemic disease, likely due to the very low concentrations of circulating lymphoma precursors (28).

In this study, we sought to directly and comprehensively profile the genetic hierarchy of precursor lesions contributing to FL. We applied ultradeep targeted genotyping (29, 30) of 334 commonly mutated lymphoma genes using Cancer Personalized Profiling by Deep Sequencing (CAPP-seq; ref. 31) to four groups of FL subjects based on tissue type. These included (i) t(14;18)-positive prediagnostic blood samples from healthy individuals who later developed FL, (ii) prediagnostic samples paired with subsequent FL tumors, (iii) paired malignant and nonmalignant lymph node biopsies, and (iv) bone marrow specimens from patients diagnosed with FL. We separately profiled blood and splenic samples from healthy adults as controls, including t(14;18)-positive and t(14;18)-undetected subjects. Our findings provide direct evidence for recurrent key FL driver mutations detectable in prediagnostic blood and tissues years prior to clinical diagnosis.

To directly profile mutations carried by t(14;18)-translocated CPC years before FL diagnosis, we performed ultra-deep targeted genotyping of 334 FL-associated genes in blood samples from 27 selected subjects from the large, population-based European Prospective Investigation into Cancer and Nutrition (EPIC) study (Fig. 1A; ref. 32). This group was selected from 318 healthy EPIC participants previously found to carry elevated t(14;18) frequencies compatible with the detection sensitivity of error-corrected CAPP-seq (6, 29). We analyzed 17 patients who developed FL [referred to as prediagnostic FL (PDFL) patients] at a median time to diagnosis of 6.3 years (range, <1–12 years).

To address the specificity of our findings, 10 EPIC subjects who remained healthy were selected as controls based on comparable t(14;18) frequencies (Fig. 1A; Supplementary Methods; Supplementary Table S1). The median age at screening was 50 years (range, 30–64) for PDFL patients and 51 years (range, 44–63) for BCL2-translocated controls. We additionally profiled an independent cohort of 20 unselected blood samples from healthy adults (median age 42.5 years) to a similar median sequencing depth (Supplementary Fig. S1; Supplementary Table S2) as well as splenic biopsies from organ donors carrying t(14;18) translocations but without evidence of lymphoma (n = 3; ref. 33).

Prediagnostic samples from patients later developing FL were significantly enriched in mutations in the CREBBP gene, seen in 7 of 17 PDFL cases (41% compared with 6.5% in controls, P = 0.004) at an average allelic fraction (AF) of 0.41% (range, 0.04%–2.2%; Fig. 1B; Supplementary Table S3). CREBBP lesions were detected at a median of 7.8 years and as early as 12 years prior to clinical diagnosis. Additional mutations were detected at a lower prevalence in other genes frequently mutated in FL: KMT2D (4/17), TP53 (2/17), BCL2 (1/17), STAT3 (2/17), and TNFRSF14 (1/17). CREBBP-mutated cases displayed a trend toward higher median t(14;18) frequency (Supplementary Fig. S2).

In the t(14;18)-positive control group including 10 blood and 3 splenic samples, coincident CREBBP and EP300 mutations were seen in 2 of 13 cases (15.4%), with both CREBBP mutations occurring outside characterized functional domains, as well as 2 KMT2D missense mutations. Among 20 healthy blood donors, we detected two variants in STAT3 and CXCR4 (Fig. 1B; Supplementary Table S4). Somatic mutations were more frequently detected in PDFL patients than t(14;18)-positive controls (82.3% vs. 60%). Both t(14;18)-positive groups had a higher mutational burden than blood donor controls (Fig. 1C), suggesting an active mutational process in translocation-bearing B cells (33). Furthermore, PDFL patients bore significantly larger clones relative to t(14;18)-translocated controls (P = 0.015; Fig. 1D). Although cases were selected for detectable t(14;18) by PCR, only one variant exceeded the clinical CHIP threshold of 2% (Fig. 1E).

To confirm that precursor variants are persistent and clonally dominant in clinical FL, we identified 34 cases with FL tumor biopsies and paired prediagnostic specimens. These included 29 individuals from the population-based American Cancer Society Cancer Prevention Study II (ACS CPS-II) with screening blood or saliva sample, subsequent FL diagnosis, and available tumor biopsy (Fig. 2A; ref. 34). The median age at screening was 71 years (range, 56–83) with a median time to FL diagnosis of 5 years (range, <1–12; Supplementary Table S1). Tumor biopsies were also available for five EPIC cases. After identifying mutations in FL tumors, we sequenced matched prediagnostic specimens via error-corrected CAPP-seq to a median 7,442× depth (Supplementary Table S2).

Utilizing a previously described Monte Carlo–based statistical test (29), we detected significant tumor-derived variants in 13 of 34 paired prediagnostic specimens (38%) at a median 5 years before clinical diagnosis (range, 1–9; Fig. 2B; Supplementary Table S5). Among detected cases, in addition to recurrent noncoding BCL2 mutations adjacent to the t(14;18) breakpoint (10/12 cases, 83%), the most frequent concordant coding lesions occurred in CREBBP (7/9 cases with mutation in tumor, 78%), BCL2 (6/9 cases, 66%), and KMT2D (4/9, 44%; Supplementary Table S6). One case (#9579) appeared to be an outlier with a median AF of 2.6% at 90 months before diagnosis. For the remaining cases, the median AF of other detected variants was 0.068%.

Screening specimens in ACS CPS-II consisted of either blood or saliva specimens, with saliva cell pellets previously noted to contain hematopoietic DNA (35). We detected tumor-confirmed variants in both saliva and blood screening specimens (Fig. 2C) although at a significantly lower median AF in saliva (Fig. 2D). Individuals with detected variants were neither older at screening, suggesting against the strictly age-related cumulative acquisition of mutations, nor significantly closer to clinical diagnosis (Fig. 2E and F).

The enrichment of CREBBP mutations in prediagnostic samples could arise from length–time bias if such cases displayed more indolent disease biology, creating a longer window for detection between mutation acquisition and clinical diagnosis. We compared the time between screening and diagnosis for FL cases in both screening cohorts with or without detected CREBBP mutation and found no significant difference (Fig. 2G).

Considering the localization of precursor CREBBP mutations within functional domains, all confirmed prediagnostic coding variants localized to the lysine acetyltransferase (KAT) domain (13/13 cases; Fig. 2H). This distribution confirms prior studies showing a strong localization of mutations in mature FL tumors to the KAT domain (68.8%, Fig. 2I; refs. 15, 19, 36–39). By contrast, two CREBBP mutations seen in BCL2-translocated control subjects occurred at I1038M and P1045S, proximal to the bromodomain and outside characterized functional domains. Most importantly, considering all prediagnostic and control samples, mutations in the CREBBP KAT domain were significantly enriched in individuals who later developed FL compared with negative controls (14/48 vs. 0/33; Fisher exact P < 0.0001). This distinction remained significant when considering only control cases with detectable t(14;18) (P = 0.03).

These findings demonstrate the patterns of pathogenic CREBBP mutation in PDFL cases compared with controls, suggesting that BCL2 translocations and CREBBP mutations constitute a recurrent model for FL precursor development, with additional events accrued over years likely necessary for overt malignant transformation.

To evaluate the persistence of precursor lesions posttreatment and benefit from higher B-cell purity, we identified eight patients who had multiple excisional biopsies for clinically significant lymphadenopathy before and after FL diagnosis and therapy. For each patient, one cryopreserved biopsy revealed nonmalignant histology, whereas another demonstrated overt FL, with a median of 2.27 years (range, 0.1–6.8 years) between specimens (Fig. 3A). The nonmalignant biopsy either preceded the FL biopsy (n = 2, median lead time 2.7 years), was roughly synchronous before treatment (n = 1, 33 days), or was obtained after treatment, offering an opportunity to assess clonal populations not eradicated by therapy (n = 5, median lead time 2.9 years). All patients were without clinical evidence of FL at the time of the nonmalignant biopsy, which in posttreatment patients was frequently obtained due to concern for recurrence. Nonmalignant histologies included reactive follicular hyperplasia, necrotizing lymphadenitis, and normal lymphoid tissue (Supplementary Table S7). All cases were without evidence of ISFN by blinded hematopathology review including BCL2 IHC (Fig. 3B).

Genotyping of these paired specimens revealed multiple recurrently mutated genes, most notably CMGs (Fig. 3C; Supplementary Table S8; refs. 15–17). As in the two screening study cohorts, both prediagnostic biopsies (patients 118 and 032) carried tumor-confirmed aberrations including clonal IGH rearrangements (Supplementary Table S9) and concordant CREBBP and IRF8 mutations in a case which was t(14;18)-negative by clinical PCR (patient 032; Supplementary Fig. S3).

Among five posttreatment biopsies, cases 102, 104, and 100 were found to share clonal t(14;18) (Fig. 3D). Both cases carrying CREBBP mutations in tumor (#102 and #104) had concordant CREBBP lesions in subsequent hyperplasia (Fig. 3E and F). Of note, the CREBBP mutation in case 102 was a stop-gain lesion that may play a similar role to the loss of function via KAT domain variants. Other coding mutations (EZH2^Y646F, KMT2D, and EP300, among others) were found only in tumor, suggesting that subsequent hyperplasias emerged from a residual BCL2-translocated population with a limited set of mutations. Although another t(14;18)-concordant patient sharing IGLL5 and BCL2 variants remained disease-free 7 years after initial diagnosis (#100), both patients with detected CREBBP later relapsed, confirming that these mutations were present in a clinically significant CPC population (Supplementary Table S7). Although the posttreatment hyperplasias lost multiple tumor mutations, they gained only rare private mutations (Supplementary Table S10). It is unclear whether these events acquired postchemotherapy reflect selective pressure or incidental passenger mutations.

Three benign biopsies, two posttreatment (#091 and #111) and one synchronous (#93), did not share t(14;18) with their malignant counterpart, suggesting an unrelated B-cell expansion. None of these patients showed CREBBP in benign biopsies or had disease progression at the last follow-up. Interestingly, although almost all pairs bore KMT2D mutations in the malignant biopsy (7/8, 87.5%), most were nonconcordant irrespective of the order of emergence. Similarly, EZH2 was detected in multiple malignant but not in benign specimens, suggesting these genes may serve as later-transforming rather than early CPC lesions. In both EPIC and paired biopsy specimens, the ratio of CREBBP relative to t(14;18) appeared subclonal in prediagnostic specimens. This ratio approached equality in the mature FL tumors, consistent with the selection and outgrowth of a CREBBP-mutated precursor population (Fig. 3G; median ratio at diagnosis 1.04; range, 0.68–1.30; P = 0.026 for difference).

Concordant clonal relationships may occur in presumed histologic precursors. In an illustrative independent case with available imaging, a patient undergoing radical prostatectomy was found to have involvement of a single pelvic lymph node with ISFN. Staging PET/CT showed no evidence of FL (Fig. 3H), and the patient was followed expectantly for several years, with no evidence of FL. Eight years after surgery, however, he presented with inguinal swelling and bilateral FDG-avid adenopathy on PET/CT. Excisional biopsy confirmed low-grade FL, and tumor sequencing revealed a CREBBP KAT domain variant. Retrospective sequencing of serial peripheral blood specimens from his initial surveillance showed detectable CREBBP R1446C at the earliest collected time point (AF range, 0.019–0.046%), rising to AF 0.082% after clinical diagnosis.

Collectively, these results reveal the presence of recurrent FL mutations in histologically benign lymph nodes and radiographically occult precursors and confirm CREBBP lesions and BCL2 translocations as both detectable before FL onset and reemerging in patients who eventually experience relapse.

The initial BCL2 translocation found in FL is thought to occur in pre–B cell precursors in the bone marrow (18). However, the high prevalence of CREBBP and KMT2D mutations raises the possibility that, analogous to myeloid CHIP mutations, these may occur in common hematopoietic precursors rather than in committed B-cell precursors in the marrow or germinal center (GC). To address this question of mutational hierarchy, we studied blood, bone marrow aspirate, and lymph node samples from seven patients with CREBBP and/or KMT2D mutations in their FL tumor (Supplementary Table S11). We analyzed flow-sorted purified hematopoietic cell populations by deep sequencing, including hematopoietic stem and progenitor cells (HSPC), mature B cells and T cells, or non–B fractions (Supplementary Tables S12 and S13).

Patient FL002, a 45-year-old man with relapsed FL, underwent genotyping of a diagnostic bone marrow biopsy at relapse. Despite the absence of morphologic involvement, a CREBBP mutation concordant with his FL tumor was detected in the bulk bone marrow aspirate. To ascertain the cell population bearing this mutation, we sorted high-purity viable cells from a marrow aspirate by FACS into mature B-cell (CD19⁺/CD20⁺/Lambda⁺), mature T-cell (CD19⁻/CD20⁻/CD5⁺), and precursor lymphoid (CD34⁺/CD20⁻) populations (Fig. 4A). Although the same CREBBP mutation was observed in the mature FL compartment in bone marrow at an AF of 40.3%, the genetic event was not detectable in other cell populations (Supplementary Table S14).

To validate this finding, a similar sorting strategy was used on viable bone marrow aspirate or peripheral blood samples from four FL patients bearing CREBBP mutations to purify HSPCs, recirculating FL, and non–B cells. In each case, CREBBP was detected in FL but absent from HSPCs (Fig. 4B; Supplementary Table S14). One patient (CIML004) also carried a characteristic KMT2D stop mutation, which was likewise absent from HSPCs. Although we cannot exclude the possibility of rare CREBBP/KMT2D-bearing HSPCs, our data are consistent with the accrual of CMG mutations later in hematopoiesis.

An atypical case sheds additional light on the timing of CREBBP mutations in FL. Patient 267 was diagnosed at age 46 with grade 1–2 FL bearing CREBBP and NRAS mutations and achieved a complete response after immunochemotherapy. Eight years into a prolonged remission, he developed cutaneous and gingival myeloid sarcoma with no radiographic or histopathologic evidence of FL. Although the CREBBP mutation was not detected in bulk marrow sequencing or cutaneous biopsies of myeloid sarcoma, concordant NRAS mutations and clonal VDJ rearrangements were seen in all three compartments (Fig. 4C). This unusual case of clonally related disease favors the occurrence of the CREBBP mutation in the committed B-cell lineage after pre-BCR rearrangements, suggesting localization later than the branch point between morphologically distinct lymphoid and myeloid tumors (ref. 40; Supplementary Fig. S4).

Together, these results suggest a hierarchy in which CREBBP mutations are preceded by BCL2 translocations—evidenced both by relative AFs and by nondetection in all tested HSPC compartments—and localize the likely cellular origin of CREBBP mutations to the committed B-cell lineage (Fig. 4D).

Our study provides a number of key insights into the lymphomagenesis of human FL. Based on blood and benign tissue samples collected years before diagnosis and after therapy, we had a unique opportunity to map the genomic landscape of FL CPCs and assess the temporal order of mutation acquisition. Prior analyses of concordant gene mutations in sequential tumor biopsies have suggested a stereotyped model for the development and progression of FL (15, 19). These findings have been corroborated in separate studies of anatomically separated FL deposits (11, 12). Collectively, these studies, using both space and time to infer hierarchy, have implicated multiple CMGs including CREBBP, KMT2D, EP300, and EZH2 as potential early lesions, with the majority of FL tumors harboring two or more CMG mutations (11, 16, 17, 19).

Our results provide direct prediagnostic evidence supporting these inferential studies and demonstrate CREBBP KAT domain mutations as prevailing precursor lesions directly detectable in 29% of prediagnostic samples up to 12 years prior to FL diagnosis. Based on three different approaches in blood and tissues, we found that BCL2 and CREBBP are essential features that may also survive immunochemotherapy, serving as a reservoir for clonal reemergence leading to eventual relapse. Our data converge toward an evolutionary history in which primary overt FL emerges from a common set of early mutations in the committed B-cell lineage, evolving and widely disseminating over many years in asymptomatic “subclinical” patients.

Although CREBBP mutations in diffuse large B-cell lymphomas are observed across the gene, prior genomic studies of FL have noted preferential localization to the KAT domain, which activates gene transcription via acetylation of enhancer regions (15, 19, 36–39). Evidence from tumor gene expression, cell culture, and animal models associates such lesions with impaired GC exit, decreased MHC class II expression, and aberrant T-cell proliferation (10, 11, 41, 42). Our three prediagnostic cohorts confirm the predominance of KAT domain loss of function, suggesting a major functional role in early lymphomagenesis potentially via dysfunctional maintenance of the GC phenotype and evasion of immune surveillance (10, 41).

Several other CMGs have been implicated in FL tumorigenesis. The KMT2D (MLL2) gene encodes a histone methyltransferase critical for tumor suppression, GC formation, and immunologic function and is mutated in most mature FL tumors (43, 44). In our study, we detected KMT2D mutations in the majority of mature tumors but less frequently in paired prediagnostic samples. The combination of high prevalence, truncal clonality, and apparent later chronicity is more suggestive of a critical cooperating or transforming event than an early precursor lesion. We speculate that mutations in distinct CMGs might engage unique gene expression programs associated with the GC reaction in human FL, with emerging evidence for such specificity emerging in murine lymphoma models (45).

Hematopoietic stem cells (HSC) are believed to be the cell of origin in several lymphoid leukemias, and mouse models have demonstrated lymphoma development with experimentally induced CREBBP lesions in HSCs (20, 46, 47). Cases of lymphomyeloid lineage switch and FL occurring after allogeneic stem cell transplant have also raised the suspicion that FL precursor mutations could occur very early in hematopoietic development (14, 40), and a CREBBP mutation was recently described in a CD34⁺/CD38⁺ lymphoid population from the bone marrow of a single FL patient (20). The analysis of early hematopoiesis in FL is also complicated by the frequent trafficking of FL founder cells from secondary lymphoid tissue to marrow-specific niches (48). However, in sorted hematopoietic cell populations from both marrow and peripheral blood, we observed CREBBP mutations in bulk B cells and isotype-specific FL lineages but not in CD34⁺/CD20⁻ precursor populations or paired lymphoid/myeloid disease. Our data therefore are not in support of HSPCs as the cell of origin or precursor reservoir in FL.

The clonal selection for CREBBP KAT mutations observed in our cases reinforces the functional importance of CREBBP loss of function but also emphasizes the primacy of BCL2 translocation, which is thought to occur due to errors in D–JH rearrangement after commitment to the pro–B or pre–B lineage in the bone marrow (2). Given that cells harboring the t(14;18) translocation in healthy individuals appear derived from the GC (3), recurrent mutations in CREBBP are likely to occur after the pre–B cell stage and potentially as late as in the GC, although this remains to be definitively established.

Limitations of our study include sample size, acknowledging the inherent rarity of paired screening and diagnostic specimens, and lack of comprehensive clinical follow-up data in population-level screening cohorts. The detection of exceedingly rare somatic variants using high-throughput sequencing presents an additional technical challenge. For example, as lymphoma precursor cells may be primarily localized in nodal tissue rather than in blood, highly sensitive strategies are required to detect early mutations: A recent study investigating lymphoid clonal hematopoiesis with a CHIP AF cutoff of 2% identified no lymphoid coding mutations in patients developing FL (28). Of note, we detected precursor mutations in the blood of PDFL patients at a median AF of 0.29%, whereas a similar study of myeloid precursors in the EPIC cohort detected a median AF of 4.3% (ref. 25; Supplementary Fig. S5). Nevertheless, additional technical refinements are likely to further enhance early detection of FL precursors.

Although CREBBP KAT mutations were frequently detected before FL diagnosis, these lesions were not universal, indicating that other genes likely serve as alternative secondary lesions. Possibilities include the acetyltransferase EP300, homologous to CREBBP and recurrently mutated in mature FL; however, in murine models loss of EP300 is associated with impaired B-cell fitness, and we observed two EP300 KAT domain mutations only in our control cohort (45). It is also unclear if CREBBP KAT domain mutations in conjunction with BCL2 translocation are sufficient for FL development. The long delay in diagnosis in our cohort suggests that additional genetic, epigenetic, or environmental factors are required. Although our cohort is focused by design on t(14;18)-translocated disease, a single BCL2-negative case (#032) suggests a potential role for CREBBP mutations in this setting as well.

Collectively, our study provides the first direct evidence for the presence of genetic alterations in circulating precursor cells years before overt FL development, dominated by CREBBP mutations within the KAT domain. Larger confirmatory studies are warranted to elucidate the hierarchy of subsequent genetic lesions and to identify additional cooperating lesions in cases without CREBBP aberrations. Given their prevalence, CREBBP KAT mutations in combination with an elevated frequency of circulating t(14;18)-translocated cells may be of utility in identifying patients at the highest risk of developing FL. Such strategies could perhaps allow the early identification of a subset of patients with limited-stage FL, where localized therapies can induce exceptionally durable remissions (49). Targeted treatment strategies with agents, including histone deacetylase inhibitors, are also under investigation for the therapy of CREBBP-mutant lymphomas (42). Conversely, in monitoring patients under treatment for FL, it is tempting to speculate whether it might be possible to identify and target residual precursors that serve as a reservoir for relapse after therapy.

Precursor mutations in FL were profiled in seven groups of individuals. EPIC FL cases included screening blood samples from 17 individuals participating in the EPIC study. These patients had confirmed t(14;18) frequencies greater than 7 × 10⁻⁵ in circulating leukocytes at screening and were later diagnosed with FL (6). From five of these individuals, we also obtained paired diagnostic FL tissue biopsies. EPIC controls included screening blood samples from 10 individuals participating in the EPIC study who had confirmed t(14;18) frequencies greater than 7 × 10⁻⁵ in circulating leukocytes at screening but had not developed lymphoma at a median clinical follow-up of >15 years (6). ACS FL cases included screening blood or saliva samples with paired FL tumor biopsies from 29 individuals participating in the ACS CPS-II LifeLink Cohort with subsequent FL diagnosis (34). Organ donor controls included splenic autopsy samples from three adult organ donors who had confirmed tissue t(14;18) frequencies greater than 5 × 10⁻⁴ but no clinical evidence of lymphoid malignancy or reactive hyperplasia (33). Healthy adult controls included blood samples from 20 healthy adult blood donors (median age 42.5, range, 26–66) from the Stanford Blood Center. Paired biopsy cases (PBC) included lymph node biopsies from eight individuals assessed at Stanford University for clinically significant lymphadenopathy. Bone marrow cases included paired bone marrow and tumor biopsies from seven FL patients, including one also included in the PBC. Samples from all cohorts were collected with written informed consent and were approved by the corresponding Institutional Review Boards in accordance with the Declaration of Helsinki. The EPIC study was approved by the review board of the International Agency for Research on Cancer and by the local institutes in the participating countries.

DNA extraction and qPCR were performed as previously described (6). DNA from five available formalin-fixed, paraffin-embedded (FFPE) tumor biopsies was extracted using the AllPrep DNA/RNA FFPE Kit (Qiagen).

Collection and initial processing of the screening specimens were as previously described (34). DNA from blood screening samples was extracted with the QIAmp DNA Micro Kit, whereas DNA from saliva screening samples was processed via the Qiagen supplementary protocol “Isolation of genomic DNA from saliva (QA19s Jul-10).” DNA from tumor FFPE samples was extracted using the DNAstorm FFPE kit (Cell Data Sciences).

Organ donor mononuclear cells were frozen immediately after surgical removal and placed at −80°C until DNA extraction. Healthy adult blood specimens were centrifuged at 1,800 × g for 10 minutes, and plasma was removed. DNA was extracted using the DNeasy Blood and Tissue Kit (Qiagen). Specimens were evaluated for detectable t(14;18) as discussed below in “Identification and Quantification of BCL2 Rearrangements” (6).

OCT samples were cut into scrolls with a cryostat, removed from the cryopreservation medium, and rinsed in a cold 70% ethanol solution prior to DNA extraction. Genomic DNA was isolated using the AllPrep DNA/RNA Mini Kit (Qiagen).

A fraction of the bone marrow aspirate was isolated with the SepMate protocol (STEMCELL) and viably stored in liquid nitrogen for subsequent FACS experiments. Paired tumor biopsy samples were preserved as FFPE for clinical pathology. All patients had CREBBP mutations identified in their FL tumor.

All biopsy samples were histologically diagnosed by pathologists according to World Health Organization criteria at the time of collection. Histologic diagnoses were confirmed by hematopathology review for all Stanford biopsy samples and additionally reviewed by a blinded hematopathologist (Y. Natkunam).

Targeted sequencing by the CAPP-seq approach was performed on all blood and tissue samples (29, 31, 50). Samples underwent hybridization capture with customized sequencing panels to enrich for commonly mutated lymphoma genes of interest (Supplementary Table S15). Stereotypical errors and PCR duplicates were removed using a nucleotide-level background database and specialized computational workflow including deduplication as previously described (29).

For diagnostic tumor samples, single-nucleotide variants and insertions/deletions were genotyped as previously described (29, 31, 50). Identified genetic aberrations are listed in Supplementary Tables S3–S5, S8, and S14. The detection of tumor-derived mutations in prediagnostic specimens was assessed for statistical significance using a previously described Monte Carlo–based test developed for tumor variant detection at low levels in cell-free DNA (29).

BCL2 translocations in the EPIC cohorts were analyzed (i) by qPCR with breakpoint sequences determined by Sanger sequencing as previously described (6), (ii) by calculating the ratio of translocated reads relative to wild-type IGH and BCL2 regions, and (iii) by using the Delly2 software package (51).

Details of FACS-sorted subpopulations in each patient sample are given in Supplementary Table S12, and details of antibodies are provided in Supplementary Table S13.

To compare continuous data distributions, we used the nonparametric Mann–Whitney U test. Two-sided Fisher exact test was used to determine whether mutations were enriched in patient groups. Mutant allele frequencies were expressed as a percentage defined as the number of reads harboring a given variant divided by the total number of reads at a given genomic position. Statistical tests were performed using GraphPad Prism version 7.0d and R version 4.0.3 with P values <0.05 considered significant. Comparisons of sets of ratios and fold changes were performed using the BootStRatio resampling method (52).

The human sequence raw data generated in this study are protected per the patient privacy policies of the original population screening studies but are available upon reasonable request from the corresponding authors subject to institutional approvals. Data derived from the EPIC study are subject to the EPIC-Europe Data and Biospecimen Access Policy, revised February 1, 2023. Data derived from the ACS CPS-II study is subject to the ACS Department of Population Science Data Access Policies and Procedures, revised December 8, 2020. Variant-level and translocation data generated in this study are available within the article and its supplementary data files. This study does not report the original code.

J. Soo reports grants from the Doris Duke Research Fellowship during the conduct of the study. F. Scherer reports grants from Roche Sequencing Solutions, Takeda, and Gilead outside the submitted work. D.M. Kurtz reports other support from Foresight Diagnostics, and personal fees from Roche and Genentech outside the submitted work, as well as a patent for U.S. Pat. No. 11,447,833 issued, licensed, and with royalties paid from Foresight Diagnostics. B.J. Sworder reports personal fees from Foresight Diagnostics outside the submitted work. M.S. Khodadoust reports grants from Nutcracker Therapeutics, and personal fees from Daiichi Sankyo and Myeloid Therapeutics outside the submitted work. Y. Natkunam reports personal fees from Roche and Leica Biosystems, and grants from Kite Pharma outside the submitted work. B. Nadel reports grants from the Fondation ARC and Agence Nationale de la Recherche during the conduct of the study, as well as grants from Bristol Myers Squibb and Roche outside the submitted work. M. Diehn reports grants and personal fees from AstraZeneca and Genentech, personal fees from Illumina, Novartis, Boehringer Ingelheim, and Bristol Myers Squibb, personal fees and other support from Gritstone Oncology, and other support from CiberMed and Foresight Diagnostics outside the submitted work, as well as a patent for circulating tumor DNA (ctDNA) methods issued, licensed, and with royalties paid from Roche, a patent for ctDNA methods issued, licensed, and with royalties paid from Foresight Diagnostics, and a patent for single-cell analysis methods issued, licensed, and with royalties paid from Celgene. A.A. Alizadeh reports other support from Foresight Diagnostics, CiberMed, and Cargo Therapeutics outside the submitted work; a patent for CAPP-seq issued, licensed, and with royalties paid from Roche; and consultancy for Celgene, Chugai, Genentech, Gilead, Janssen, Pharmacyclics, and Roche, scientific advisory board membership in the Lymphoma Research Foundation, professional affiliations with the American Society of Hematology, the American Society of Clinical Oncology, the American Society of Clinical Investigation, and The Leukemia & Lymphoma Society, research funding from the NCI, the National Heart, Lung, and Blood Institute, the NIH, Celgene, Bristol Myers Squibb, and Pfizer, patent filings including patent issued, licensed, and with royalties paid from FortySeven, a patent pending and licensed to Foresight Diagnostics, a patent pending relating to MARIA, a patent issued and licensed to CiberMed, a patent issued, a patent pending to CiberMed, patents issued to Idiotype Vaccines, and a patent issued, licensed, and with royalties paid from Roche, and equity ownership interests in CiberMed, Foresight Diagnostics, FortySeven, and Cargo Therapeutics. No disclosures were reported by the other authors.

J.G. Schroers-Martin: Conceptualization, data curation, formal analysis, investigation, visualization, writing–original draft, writing–review and editing. J. Soo: Investigation, methodology, writing–original draft, writing–review and editing. G. Brisou: Resources, validation, investigation, methodology. F. Scherer: Investigation, visualization, writing–original draft, writing–review and editing. D.M. Kurtz: Investigation, writing–review and editing. B.J. Sworder: Investigation, methodology, writing–review and editing. M.S. Khodadoust: Investigation, methodology. M.C. Jin: Investigation. A. Bru: Investigation. C.L. Liu: Resources, data curation, software. H. Stehr: Data curation, software. P. Vineis: Resources, supervision, project administration. Y. Natkunam: Resources, writing–review and editing. L.R. Teras: Resources, validation. J.Y. Song: Resources, data curation, investigation, writing–review and editing. B. Nadel: Conceptualization, resources, supervision, writing–review and editing. M. Diehn: Conceptualization, supervision, funding acquisition, project administration, writing–review and editing. S. Roulland: Conceptualization, resources, supervision, investigation, methodology, writing–original draft, project administration, writing–review and editing. A.A. Alizadeh: Conceptualization, resources, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

This study was financially supported by NCI R01CA233975 (A.A. Alizadeh and M. Diehn); NIH 1U01 CA194389 (A.A. Alizadeh); the NIH Director's New Innovator Award Program (1-DP2-CA186569 to M. Diehn); the Virginia and D.K. Ludwig Fund for Cancer Research (A.A. Alizadeh and M. Diehn); a Lymphoma Research Foundation FL Pathways Award (A.A. Alizadeh); the Damon Runyon Cancer Research Foundation (A.A. Alizadeh and D.M. Kurtz); an American Society of Hematology Scholar Award (A.A. Alizadeh); The Leukemia & Lymphoma Society (A.A. Alizadeh); the V Foundation (A.A. Alizadeh); a Stanford Cancer Innovation Award (A.A. Alizadeh); the ARC Foundation (B. Nadel); Institut Carnot CALYM (B. Nadel); Plan Cancer Epigenetique et Cancer (S. Roulland); a Canceropole PACA Grant (S. Roulland and B. Nadel); and Fondation pour la Recherche Médicale (G. Brisou). Philanthropic support was provided by the Moghadam Family Endowed Professorship (A.A. Alizadeh); The CRK Faculty Scholar Fund (M. Diehn); the German Research Foundation (F. Scherer, SCHE 1870/1-1); Stanford Cancer Institute (J.G. Schroers-Martin and A.A. Alizadeh); the Doris Duke Charitable Foundation (J. Soo and A.A. Alizadeh); the Bakewell Foundation (A.A. Alizadeh and M. Diehn); SDW/DT and Shanahan Family Foundations (A.A. Alizadeh); The Skeff Family Lymphoma Fund (A.A. Alizadeh); the Jewish Communal Fund for Lymphoma Research (A.A. Alizadeh); The Arzang Family Lymphoma Fund (A.A. Alizadeh); The Cane-Nowak Family Foundation (A.A. Alizadeh); The Troper-Wojcicki Family Gift (A.A. Alizadeh); The Marc Benioff Fund (A.A. Alizadeh); The Sara Schottenstein Memorial Fund (A.A. Alizadeh); and anonymous philanthropic donors (A.A. Alizadeh). We are grateful to the participants in the EPIC and ACS CPS-II studies and to the patients at Stanford University and the hematology department of Lyon-Sud Hospital for their participation. We are grateful for the technical assistance of the EPIC centers in Spain and Italy in providing tumor biopsies, and the insightful discussions with the members of the Alizadeh and Diehn laboratories.